doi
stringlengths 10
10
| chunk-id
stringlengths 1
4
| chunk
stringlengths 1
1.66k
| id
stringlengths 10
10
| title
stringlengths 19
148
| summary
stringlengths 345
1.92k
| source
stringlengths 31
31
| authors
list | categories
list | comment
stringlengths 4
284
⌀ | journal_ref
stringclasses 14
values | primary_category
stringclasses 16
values | published
stringlengths 8
8
| updated
stringlengths 8
8
| references
list |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
1801.04871
|
35
|
meal=dinner, location=near the
theatre)Find restaurant with meal is dinner
and location is near the theatre.I want to get dinner at a restaurant
near the theatre.
S: request(cuisine, price range) Provide cuisine and price range.Do you have any preference for
cuisine or price range?
U: inform(cuisine=DontCare,
price range=moderate, rating=high)Cuisine is I don’t care and price
range is moderate and rating is high.I’m fine with any cuisine. Look for
something moderately priced, but
make sure it has a high rating.
S: select(restaurant= fFirst Wok,
Lucy’s Grillg, location=near the
theatre)Select restaurant from First Wok,
Lucy’s Grill with location is near the
theatre.First Wok and Lucy’s Grill are
some good options near the
theatre.
U: inform(intent=reserve restaurant,
restaurant=First Wok, time=after the
movie)Reserve restaurant with restaurant is
First Wok and time is after the
movie.First Wok sounds perfect. Can you
reserve a table there for dinner
after the movie?
S: ack() confirm(restaurant=First
|
1801.04871
|
Building a Conversational Agent Overnight with Dialogue Self-Play
|
We propose Machines Talking To Machines (M2M), a framework combining
automation and crowdsourcing to rapidly bootstrap end-to-end dialogue agents
for goal-oriented dialogues in arbitrary domains. M2M scales to new tasks with
just a task schema and an API client from the dialogue system developer, but it
is also customizable to cater to task-specific interactions. Compared to the
Wizard-of-Oz approach for data collection, M2M achieves greater diversity and
coverage of salient dialogue flows while maintaining the naturalness of
individual utterances. In the first phase, a simulated user bot and a
domain-agnostic system bot converse to exhaustively generate dialogue
"outlines", i.e. sequences of template utterances and their semantic parses. In
the second phase, crowd workers provide contextual rewrites of the dialogues to
make the utterances more natural while preserving their meaning. The entire
process can finish within a few hours. We propose a new corpus of 3,000
dialogues spanning 2 domains collected with M2M, and present comparisons with
popular dialogue datasets on the quality and diversity of the surface forms and
dialogue flows.
|
http://arxiv.org/pdf/1801.04871
|
[
"Pararth Shah",
"Dilek Hakkani-Tür",
"Gokhan Tür",
"Abhinav Rastogi",
"Ankur Bapna",
"Neha Nayak",
"Larry Heck"
] |
[
"cs.AI",
"cs.CL"
] |
11 pages, 4 figures
| null |
cs.AI
|
20180115
|
20180115
|
[
{
"id": "1801.04871"
},
{
"id": "1709.00103"
},
{
"id": "1712.01815"
},
{
"id": "1705.06476"
},
{
"id": "1704.00057"
}
] |
1801.04871
|
36
|
movie.First Wok sounds perfect. Can you
reserve a table there for dinner
after the movie?
S: ack() confirm(restaurant=First
Wok, time=8pm, num people=2)OK. Confirm restaurant is First Wok
and time is 8pm and num people is 2.Sure. Please confirm that it is a table
for 2 at First Wok for 8pm.
U: affirm() Agree. That is correct.
S: notify success() Reservation confirmed. Your table has been reserved.
U: thank you() good bye() Thank you and good bye. Thanks! That’s all for now.
Figure 3: Contextual rewrite task interface for paraphrasing a dialogue outline with natural language.
|
1801.04871
|
Building a Conversational Agent Overnight with Dialogue Self-Play
|
We propose Machines Talking To Machines (M2M), a framework combining
automation and crowdsourcing to rapidly bootstrap end-to-end dialogue agents
for goal-oriented dialogues in arbitrary domains. M2M scales to new tasks with
just a task schema and an API client from the dialogue system developer, but it
is also customizable to cater to task-specific interactions. Compared to the
Wizard-of-Oz approach for data collection, M2M achieves greater diversity and
coverage of salient dialogue flows while maintaining the naturalness of
individual utterances. In the first phase, a simulated user bot and a
domain-agnostic system bot converse to exhaustively generate dialogue
"outlines", i.e. sequences of template utterances and their semantic parses. In
the second phase, crowd workers provide contextual rewrites of the dialogues to
make the utterances more natural while preserving their meaning. The entire
process can finish within a few hours. We propose a new corpus of 3,000
dialogues spanning 2 domains collected with M2M, and present comparisons with
popular dialogue datasets on the quality and diversity of the surface forms and
dialogue flows.
|
http://arxiv.org/pdf/1801.04871
|
[
"Pararth Shah",
"Dilek Hakkani-Tür",
"Gokhan Tür",
"Abhinav Rastogi",
"Ankur Bapna",
"Neha Nayak",
"Larry Heck"
] |
[
"cs.AI",
"cs.CL"
] |
11 pages, 4 figures
| null |
cs.AI
|
20180115
|
20180115
|
[
{
"id": "1801.04871"
},
{
"id": "1709.00103"
},
{
"id": "1712.01815"
},
{
"id": "1705.06476"
},
{
"id": "1704.00057"
}
] |
1801.04871
|
37
|
Figure 4: Dialogue quality evaluation task interface for rating the user and system turns of completed dialogues.
|
1801.04871
|
Building a Conversational Agent Overnight with Dialogue Self-Play
|
We propose Machines Talking To Machines (M2M), a framework combining
automation and crowdsourcing to rapidly bootstrap end-to-end dialogue agents
for goal-oriented dialogues in arbitrary domains. M2M scales to new tasks with
just a task schema and an API client from the dialogue system developer, but it
is also customizable to cater to task-specific interactions. Compared to the
Wizard-of-Oz approach for data collection, M2M achieves greater diversity and
coverage of salient dialogue flows while maintaining the naturalness of
individual utterances. In the first phase, a simulated user bot and a
domain-agnostic system bot converse to exhaustively generate dialogue
"outlines", i.e. sequences of template utterances and their semantic parses. In
the second phase, crowd workers provide contextual rewrites of the dialogues to
make the utterances more natural while preserving their meaning. The entire
process can finish within a few hours. We propose a new corpus of 3,000
dialogues spanning 2 domains collected with M2M, and present comparisons with
popular dialogue datasets on the quality and diversity of the surface forms and
dialogue flows.
|
http://arxiv.org/pdf/1801.04871
|
[
"Pararth Shah",
"Dilek Hakkani-Tür",
"Gokhan Tür",
"Abhinav Rastogi",
"Ankur Bapna",
"Neha Nayak",
"Larry Heck"
] |
[
"cs.AI",
"cs.CL"
] |
11 pages, 4 figures
| null |
cs.AI
|
20180115
|
20180115
|
[
{
"id": "1801.04871"
},
{
"id": "1709.00103"
},
{
"id": "1712.01815"
},
{
"id": "1705.06476"
},
{
"id": "1704.00057"
}
] |
2205.11503
|
0
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary
Textual Style Transfer with Small Language Models
Mirac Suzgun*
Stanford University
[email protected] Melas-Kyriazi*
Oxford University
[email protected] Jurafsky
Stanford University
[email protected]
Abstract
We propose a method for arbitrary textual
style transfer (TST)—the task of transforming
a text into any given style—utilizing generalpurpose pre-trained language models. Our
method, Prompt-and-Rerank , is based on a
mathematical formulation of the TST task, decomposing it into three constituent components: textual similarity ,target style strength ,
and fluency . Specifically, our method first
uses zero-shot or few-shot prompting to obtain a set of candidate generations in the target style, and then re-ranks these candidates
according to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform
on par with state-of-the-art large-scale models
while consuming two orders of magnitude less
compute and memory. Finally, we conduct a
systematic investigation of the effect of model
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
1
|
while consuming two orders of magnitude less
compute and memory. Finally, we conduct a
systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on style
transfer quality across seven diverse textual
style transfer datasets.1
1 Introduction
Textual style transfer (TST) refers to the task of
transferring one stylistic aspect of a piece of text
(e.g., sentiment attribute, formality, politeness, etc.)
without changing its main semantic content, structure, or other attributes. Traditionally, the natural language generation (NLG) community has approached each instantiation of style transfer as a
distinct task, designing and training specialized
models on style-specific training corpora. For example, sentiment transfer has been studied extensively (Li et al. (2018); Sudhakar et al. (2019);
Luo et al. (2019a), inter alia ). This paradigm has
restricted TST research to a limited, simple set of
style choices with parallel corpora, which can often
be solved adequately by word replacement (e.g., replacing negative words with corresponding positive
words for sentiment transfer).
1Our code, data, and results are available at https://
github.com/suzgunmirac/prompt-and-rerankSentiment Transfer
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
2
|
words for sentiment transfer).
1Our code, data, and results are available at https://
github.com/suzgunmirac/prompt-and-rerankSentiment Transfer
Input the food was low quality and thrown together quickly with little flavor. [Neg]
GPT-J-6B the food was delicious and prepared with a lot of care! [Pos]
Input the staff is awesome and location is right in the heart of old town! [Pos]
GPT-J-6B the staff is terrible and location is in the middle of nowhere! [Neg]
Elizabethan!Modern English Translation
Inputthou hast quar rell’d with a man for coughing in the street, because he hath
wakened thy dog that hath lain asleep in the sun. [Elizabethan]
GPT-J-6Byou have cursed a man because he’s coughed in the street, because he has woken
your dog, which has been sleeping in the sun. [ModernEng]
Input if thou art fickle, what dost thou with him that is renown’d for faith? [Elizabethan]
GPT-J-6Bif you are fickle, then what are you doing with that person who is known for
being reliable? [ModernEng]
Fluency and Grammar Correction
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
3
|
being reliable? [ModernEng]
Fluency and Grammar Correction
Input Forexample, My cousin is 12years old. [Ungrammatical]
GPT-J-6B For example, My cousin is twelve years old. [Grammatical]
Formality Transfer
Input r u talking about ur avatar? [Informal]
GPT-J-6B are you talking about your avatar? [Formal]
Table 1: Qualitative examples of few-shot style transfer on the
YELP,SHAKESPEARE ,JFLEG , and GYAFC datasets. Coupling off-the-shelf “small” language models with our promptand-reranking method enables us to perform arbitrary textual
style transfer without any model training or prompt-tuning.
Compared to the extremely large language models (viz., ones
with more than 100 billion parameters) used by Reif et al.
(2022), our models obtain similar performance using almost
two orders of magnitude less compute and memory.
With the recent success of general-purpose language modeling (LM), it is, however, natural to ask
whether one can tackle a more general formulation
of style transfer: arbitrary TST, in which one aims
to transform a reference text into an arbitrary style
specified by the user at inference-time.
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
4
|
of style transfer: arbitrary TST, in which one aims
to transform a reference text into an arbitrary style
specified by the user at inference-time.
Inspired by the success of natural-language
prompting in other domains (Radford et al., 2019;
Petroni et al., 2019; Brown et al., 2020; Gao et al.,
2021), we consider a prompting-based zero- and
few-shot approach to arbitrary TST. Under this
setup, we specify the desired type of style transfer
problem using a natural-language prompt containing the source text (and optionally a few examples,
in the few-shot case), and then use a pre-trained LM
to generate the stylized target text. Thus, the source
text may be transformed into any user-specified
style without additional training or fine-tuning.
Recent work (Reif et al., 2022) has found that extremely large language models (LLMs), namely the
175 billion-parameter GPT-3 (Brown et al., 2020)arXiv:2205.11503v1 [cs.CL] 23 May 2022
model and the proprietary 137 billion-parameter
LLM model, are capable of sentiment and formality
transfer. However, language models at this scale
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
5
|
model and the proprietary 137 billion-parameter
LLM model, are capable of sentiment and formality
transfer. However, language models at this scale
are not accessible to most researchers and practitioners, even in inference-only settings, due to their
large memory consumption and slow generation
times. Thus far, to the best of our knowledge, there
has not been any research on the capabilities of
reasonably-sized models for style transfer domain,
nor any systematic study of how the precise construction of the prompt affects model performance.
Differently from past work, this present paper
takes a first-principles approach to arbitrary TST
using pretrained language models. We first mathematically formalize the task, showing how it can be
formulated as the combination of textual similarity ,
target style strength , and fluency . This framework
naturally leads us to propose a new method for arbitrary TST, which we call “ Prompt-and-Rerank .”
Using this method, we demonstrate, for the first
time, that it is possible to perform arbitrary TST
using small language models; prior work indicated
that only enormous (i.e., GPT-3-scale) language
models were capable of this task.
We summarize the main contributions and insights of this paper as follows: (i) We provide the
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
6
|
models were capable of this task.
We summarize the main contributions and insights of this paper as follows: (i) We provide the
first mathematical formalization of the arbitrary
TST task. (ii) We propose Prompt-and-Rerank, a
novel prompting-based method for arbitrary TST
which follows naturally from our mathematical formulation. (iii) Our method matches and sometimes
even exceeds state-of-the-art performance on arbitrary TST while using small language models such
as GPT-2, which consume two orders of magnitude less memory and compute than prior work.
(iv) We conduct a nuanced investigation of the influence of prompt design, such as task phrasing
and delimiter-pair choice, on the quality of style
transfer generations. (v) In order to encourage and
facilitate further research in the area, we establish
a set of benchmarks for arbitrary TST (including
cleaned versions of the popular sentiment transfer
datasets AMAZON andYELP) along with accompanying automatic evaluation metrics.
2 Background and Related Work
Background. TST is a long-standing problem in
NLP which encompasses many popular sub-tasks,
such as sentiment and formality transfer. Prior
to the advent of large-scale pre-training in recent
years, it was common practice to consider each
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
7
|
such as sentiment and formality transfer. Prior
to the advent of large-scale pre-training in recent
years, it was common practice to consider each
of these sub-tasks separately, and to train sepa-rate models on different supervised datasets for
each task. These models generally performed well
within the limited scope of their task, but failed to
generalize to new tasks or to texts outside of their
training distribution. Here we show that the modern paradigm of pre-training large models and then
prompting (or fine-tuning) them can be applied to
many sub-tasks of TST in a unified, zero-shot manner, even with relatively small Transformers.
Related Work. Traditional approaches to TST
can be broadly categorized into two families. The
first family involves identifying and replacing distinctive style-related phrases (Li et al. (2018); Sudhakar et al. (2019); Wu et al. (2019); Madaan
et al. (2020); Malmi et al. (2020); Reid and Zhong
(2021), inter alia ). For example, Madaan et al.
(2020) perform the task of politeness transfer by
first identifying words with stylistic attributes using TF-IDF and then training a model to replace
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
8
|
(2020) perform the task of politeness transfer by
first identifying words with stylistic attributes using TF-IDF and then training a model to replace
or augment these stylistic words with ones associated with the target attribute. In general, these
approaches perform well for very simple style edits
(e.g., negating a sentence by adding the word not),
but they struggle in scenarios that require more
complex syntactic and semantic changes.
The second family of approaches involves disentangling latent representations of style and content, such that a text can be encoded into a styleinvariant representation and then decoded in a desired style (Hu et al., 2017; Shen et al., 2017; Fu
et al., 2018; Luo et al., 2019a). For example, Hu
et al. (2017) encodes into and decodes from a styleagnostic latent space using a V AE alongside attribute discriminators. These approaches are often
theoretically well-grounded, but they generally require large quantities of labeled data and struggle
to scale beyond a small number of styles.
Differently from these two families, one recent
work (Reif et al., 2022) uses enormous pre-trained
language models to tackle TST, an idea motivated
by the remarkable performance of pre-trained LMs
in other areas of NLP (Radford et al., 2019; Devlin
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
9
|
language models to tackle TST, an idea motivated
by the remarkable performance of pre-trained LMs
in other areas of NLP (Radford et al., 2019; Devlin
et al., 2019; Yang et al., 2019; Liu et al., 2019).
Specifically, they use LLM ,LLM-Dialog , and GPT3, each of which has over 100 billion parameters,
to rewrite texts in a variety of styles. However, they
perform minimal analysis of their prompting setup,
deferring such analysis to future work, and they
suggest that this prompting-based approach is only
feasible with LLMs.2
2A note on terminology : We shall refer to GPT-3 (Brown
2
This paper presents a novel prompt-and-rerank
approach to the general task of textual style transfer
using pre-trained language models. Alongside our
method, we present the first systematic study of
prompt formulation and model size for the task
of textual style transfer. Contrary to expectations,
using our method we find that even small LMs are
able to effectively perform arbitrary style transfer.
In fact we match the performance of Reif et al.
(2022) on multiple datasets using two orders of
magnitude less memory and compute.
3 Method: Prompt-Based Arbitrary TST
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
10
|
(2022) on multiple datasets using two orders of
magnitude less memory and compute.
3 Method: Prompt-Based Arbitrary TST
This section begins with a mathematical formalization of the task of textual style transfer.3Our
formalization elucidates the three underlying components of the task, namely text similarity ,target
style strength , and fluency , and naturally leads us to
Prompt-and-Rerank , our prompt-based re-ranking
algorithm for solving TST.
3.1 Problem Formulation
Letx2denote a text over a vocabulary , and
Sthe set of all possible text style choices. Let us
further use x(s1)2to denote a text xwritten
in the styles12S. Informally speaking, the goal
of TST is to transfer the style of a text x(s1)(usually, a sentence) from s1tos2without changing
the main semantic content of the text. We can formally express this transformation via a function
f: SS! , which takes an input text
(sayx(s1)) and its corresponding style ( s1), as well
as a target style ( s2), and outputs a modified version of the input written in the style of s2(namely,
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
11
|
as a target style ( s2), and outputs a modified version of the input written in the style of s2(namely,
~x(s2)).4Ideally, we would want the generated output~x(s2)=f(x(s1);s1;s2)to be “close” (both
semantically and syntactically) to the ground-truth
x(s2)as much as possible.
The graphical models depicted in Figure 1 provide two different ways of formulating the task of
et al., 2020) and similar models with 100 billion model parameters as large language models (LLMs), and the versions
of GPT-2 and GPT-J—ranging from OpenAI’s GPT-2-Small
(117M) to EleutherAI’s GPT-J-6B—which are two-to-three
orders of magnitude smaller than GPT-3, as small language
models (SLMs).
3Despite its important role in NLG, we are not aware of
any prior formal statement of the (textual) style transfer problem. Here, we hope to solidify the problem formulation and
illustrate the a connection between this problem formulation
and the automatic metrics used in the field to evaluate TST
models.
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
12
|
illustrate the a connection between this problem formulation
and the automatic metrics used in the field to evaluate TST
models.
4In cases where the original style of the input text might
not be known a priori, one can either estimate the style of the
input using a statistical classifier or assume that the input is
written in a neutral style.xs2 s1
~x(s2)x(s1)xs2 s1
~x(s2)x(s1)
Figure 1: Two different but equally meaningful and valid
interpretations of the textual style transfer task. Here xcan be
thought as the universal (abstract) meaning of a text, x(s1)a
rewrite of xin the style of s1. Depending on which graphical
model one adheres to, x(s2)can be said to generated by xand
s2(left model) or by x(s1)ands2(right model). In this paper,
we follow the second interpretation.
TST (and of machine translation for that matter).
Both models have valid and meaningful implications and interpretations; the main generative difference between them is that the parents of ~x(s2)are
xands2in the former (left), whereas the parents
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
13
|
xands2in the former (left), whereas the parents
of~x(s2)arex(s1)ands2in the latter (right).
Due to the inherent difficulty of collecting diverse supervised data for arbitrary TST, most prior
studies considered a simplified version of the task,
wherein the source ( s1) and target (s2) style choices
are fixed beforehand. In this work, we consider a
broad formulation of the task, make no assumptions about the source and target style choices a
priori, and explain how one can leverage the power
of off-the-shelf LMs to perform arbitrary TST.
Given an input text x(s1)written in the style
ofs1and the target style s2, we decompose the
conditional likelihood of a generated output ~x(s2)
into three terms:5
p(~x(s2)j[x(s1);s1];s2) (1)
=p(~x(s2);[x(s1);s1];s2)
p([x(s1);s1];s2)
/p([x(s1);s1];[~x(s2);s2])
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
14
|
p([x(s1);s1];s2)
/p([x(s1);s1];[~x(s2);s2])
=p([x(s1);s1]j[~x(s2);s2])p([~x(s2);s2])
=p([x(s1);s1]j[~x(s2);s2])| {z }
textual similarityp(s2j~x(s2))| {z }
transfer strengthp(~x(s2))|{z}
fluency
The first term, p([x(s1);s1]j[~x(s2);s2]), can be
thought as a measure of textual similarity between
the input text and the generated output. The second
term,p(s2j~x(s2)), measures the transfer strength
of the output (i.e., determines whether the output is
written in the target style). The last term, p(~x(s2)),
measures the overall fluency of the output.
5We make use of the brackets “ []” only to group relevant
terms (e.g., x(s1);s1) together; they do not have any statistical
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
15
|
terms (e.g., x(s1);s1) together; they do not have any statistical
significance in this context.
3
Output-1Output-2Output-3Output-4Output-5PromptLM(GPT-2)
BERTScoreTextual SimilarityMLM (RoBERTa)Style ClassifierGPT-2-LargeFluencyxxxxxxxxxxRe-score and re-rank the outputs.Pick the highest scoring output.Given the input text, source style, andtarget style, write a proper prompt.
Generate multiple outputsusing beam search.Output-3InputSource StyleTarget StyleFigure 2: An illustration of our Prompt-and-Rerank method. Given an input text and the style transformation, we first compose
a prompt and feed it to a pretrained language model to generate multiple output texts—conditioned on the prompt—using beam
search. We then re-score each candidate output along the three axes from Eq. (1): textual similarity ,style transfer strength , and
fluency . We choose the candidate with the highest re-ranked score as our final output.
3.2 Prompt-and-Rerank for Arbitrary TST
The problem formulation above naturally leads us
to a method for (textual) style transfer, which we
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
16
|
3.2 Prompt-and-Rerank for Arbitrary TST
The problem formulation above naturally leads us
to a method for (textual) style transfer, which we
denote Prompt-and-Rerank (P&R).
The foundation of our method is use of prompt
templates to convert TST into a natural-language
generation problem. Formally, we use a predefined
template 2T to convert an input text x(s1)and
the desired style transformation (i.e., s1!s2) into
a natural-language prefix (x;s1;s2). The template serves to not only contextualize the task for
the model but also incorporate all the necessary conditional information (that is, input sentence, source
style, and target style) in the input context. The
precise design and composition of the templates is
the topic of the following section (§4.2).6
Next, we feed the prompt into a pre-trained LM
(e.g., GPT-2) and use the model to generate kdifferent outputs ~x(s2)
kconditioned on the prompt, each
sampled independently without updating any parameters of the model. These outputs are taken to
be our candidate outputs for re-ranking.
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
17
|
kconditioned on the prompt, each
sampled independently without updating any parameters of the model. These outputs are taken to
be our candidate outputs for re-ranking.
Finally, we re-rank our kcandidate outputs according to the decomposition in Equation 1:
preranking (~x(s2)
ij[x(s1);s1];s2) (2)
/p([x(s1);s1]j[~x(s2)
i;s2])p(s2j~x(s2)
i)p(~x(s2)
i):
And finally, we pick the output ~x(s2)
i2~Xwith
the highest re-ranking score. (Figure 2 provides an
abstract illustration of our re-ranking algorithm.)
All that remains is to describe how to calculate
each term in the re-ranking pass. (i) To calculate the first (textual similarity) term, we avail ourselves of BERTScore (Zhang et al., 2020), which
6Additionally, in the few-shot case, where we have a number of few-shot exemplars, we convert these exemplars into
meaningful prompts using the same template structure and
prepend them to the main prompt.utilizes pre-trained contextual embeddings from
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
18
|
meaningful prompts using the same template structure and
prepend them to the main prompt.utilizes pre-trained contextual embeddings from
BERT to measure the cosine similarity between
two texts.7We presume p([x(s1);s1]j[~x(s2);s2]) =
BERTScore (x(s1);x(s2)). (ii) In order to calculate the second term, we deliberately turn a masked
LM (MLM), in our case a pre-trained RoBERTa
model, into a style classifier as follows: Given
~x(s2)
i2~X(s2)andS=fs1;s2g, we convert ~x(s2)
i
into a “fill-in-the-blank” cloze statement via a predefined cloze template, that is, we rewrite it as
“The following text is <mask> :[~x(s2)
i].” We then
query the MLM to predict the masked token,8but
instead of looking at the probability distribution
over the original model vocabulary, we restrict our
attention to the elements in Sand thus consider
the likelihood of the missing token being s1or
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
19
|
instead of looking at the probability distribution
over the original model vocabulary, we restrict our
attention to the elements in Sand thus consider
the likelihood of the missing token being s1or
s2. We then normalize these probabilities by l1normalization and get a proper probability distribution forp(s2j~x(s2)).9(iii) As for the last term, we
use GPT-2-Large (774M) to determine the overall
likelihood of each candidate text.10(iv) Afterwards,
we compute the score for each candidate by multiplying (i), (ii), and (iii) accordingly; re-rank all the
candidates; and pick the one with the highest score
as the final output.11
Overall, our approach is model-agnostic, allow7Let us note that BERTScore is a symmetric function, i.e.,
BERTScore (x(s1);x(s2)) = BERTScore (x(s2);x(s1)).
Furthermore, we acknowledge that BERTScore technically
neglects the additional style information, but we believe that
this is a reasonable simplification under our framework.
8One limitation of this framework is that it assumes the
styles are associated with distinct tokens in the vocabulary.
9Of course, a more sophisticated normalization technique
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
20
|
8One limitation of this framework is that it assumes the
styles are associated with distinct tokens in the vocabulary.
9Of course, a more sophisticated normalization technique
can be employed in this setup, but this basic normalization
method seemed to be sufficient in our experiments.
10Given a text x:=x1:tof lengtht, we calculate its probability under a model asp(x) =Qt
i=1p(xijx<i)
11Since the calculation of (iii) penalizes long sequences
or sequences involving rare words, we also consider the reranking method in which we ignore the fluency factor, assuming that the sentences generated by the models are always
fluent, which, we are aware that, is a faulty assumption.
4
ing pre-trained LMs to be used out-of-the-box. Furthermore, our experiments show that with welldesigned prompts, one does notneed a massive
language model for this approach to be successful.
4 Prompt Construction
In practice, we found the specific syntax and semantics of the prompt template significantly impact
model performance. Thus, we conducted a systematic investigation of the impact of different prompt
design choices on the quality of TST generations.
4.1 Delimiter-Pairs
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
21
|
model performance. Thus, we conducted a systematic investigation of the impact of different prompt
design choices on the quality of TST generations.
4.1 Delimiter-Pairs
We experimented with ten different text boundary markers (delimiter pairs), which may be divided into two categories: those whose opening
and closing markers are identical (known as indistinguishable delimiters), and those whose markers
are different (known as complementary delimiters).
Specifically, we considered two indistinguishable
pairs (viz., quotes and dashes) and eight complementary pairs: (1) curly brackets fg, (2) square
brackets [], (3) angle brackets hi, (4) parentheses
(), (5) quotes "", (6) dashes ––, (7) triple angle
bracketshhhiii , (8) bracket quotes i"", (9) asterisk quotes * "", and (10) double curly bracket
ffgg .12In their experiments, Reif et al. (2022)
uses only curly brackets.13
4.2 Prompt Phrasing
We considered four manually-written template formatst2T for our discrete prompts:
(a)Vanilla : “Here is a text: [d1][x(s1)][d2]Here
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
22
|
(a)Vanilla : “Here is a text: [d1][x(s1)][d2]Here
is a rewrite of the text, which is [s2]:[d1]”,
(b)Contrastive : “Here is a text, which is [s1]:
[d1][x(s1)][d2]Here is a rewrite of the text, which
is[s2]:[d1]”,
(c)Negation-v1 : “Here is a text, which is [s1]:
[d1][x(s1)][d2]Here is a rewrite of the text, which
is not [s1]:[d1]”, and
(d)Negation-v2 : “Here is a text, which is not
[s2]:[d1][x(s1)][d2]Here is a rewrite of the text,
which is [s2]:[d1]”.
12We use (8), (9), and (10) to emulate blockquotes, bullet
points, and liquid tags in Markdown, respectively.
13We hypothesized that the complementary delimiter-pairs
might yield better results than the indistinguishable ones, since
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
23
|
points, and liquid tags in Markdown, respectively.
13We hypothesized that the complementary delimiter-pairs
might yield better results than the indistinguishable ones, since
it is categorically easier for models to distinguish and understand where sentences start and end. We also speculated
that delimiter-pairs that were more likely to be used as textseparators in the training data in various contexts (e.g., in code
snippets) might yield better results.Note that [d1]and[d2]denote the opening and
closing elements of the chosen delimiter-pair, respectively. In their experiments, Reif et al. (2022)
exclusively made use of the vanilla setting, which
only specifies the target style ( s2) in the second
half of the prompt; however, we initially speculated that providing useful information about the
source style ( s1) and creating a clear contrast between the source and target styles in the prompt
semantics might help pre-trained LMs to have a
better understanding of the underlying nature of
the task and improve their performance; hence, we
decided to look at the contrastive setting as well.
As for the other two negation templates, we wanted
to test how specifying the source style as the negation of the target style (viz., s1:=“nots2”) and vice
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
24
|
to test how specifying the source style as the negation of the target style (viz., s1:=“nots2”) and vice
versa might affect the model performance.14
Example. Finally, to make our prompting setup
more concrete, let us give a concrete and brief example of how we formulate a prompt. We consider
thecontrastive template with curly brackets as our
delimiter. If we have an input sentence x(s1)=“I
love The Sound of Music ; it is the best movie ever!!”
withs1=positive ands2=negative , then the prompt
under this template would be “Here is a text, which
is positive: {I love The Sound of Music ; it is the
best movie ever!!} Here is a rewrite of the text,
which is negative: {” The language model would
then generate an output by autoregressively decoding after the last delimiter, to produce a sentence
such as: “I hate The Sound of Music ; it is the worst
movie ever!!}”15
4.3 Zero-Shot vs. Few-Shot Settings
In recent years, LLMs, such as GPT-3, have proven
themselves to be resourceful few-shot learners. In
a few-shot learning setting, a model is presented
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
25
|
themselves to be resourceful few-shot learners. In
a few-shot learning setting, a model is presented
with a small set of illustrative examples, oftentimes
along with a natural-language prompt describing
the task, and expected to understand the underlying
task and make accurate predictions without performing any gradient updates to the weights of the
model at inference time. We wanted to explore
how the number of demonstrations affects the performance of our models. To that end, we also tested
the performances of our models under the zero-shot
and four-shot settings.
14The last two formats might be useful especially when we
do not have access to either the source or the target style.
15Table 8 in the Appendix provides a complete set of examples of prompts used in each task.
5
Dataset Styles Example Sentence-Pairs Test Set Size
Yelp Restaurant Reviews Negative ever since joes has changed hands it’s just gotten worse and worse.1000(Zhang et al., 2015) Positive ever since joes has changed hands it’s gotten better and better.
Amazon Product Reviews Negative if your bike had a kickstand on the plate it won’t lock down.1000(He and McAuley, 2016) Positive if your bike had a kickstand on the plate it would lock down.
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
26
|
GYAFC Formality Dataset Informal and so what if it is a rebound relationship for both of you?1000(Rao and Tetreault, 2018) Formal what if it is a rebound relationship for both of you?
Shakespearean English Dataset Elizabethan is rosaline, whom thou didst love so dear, so soon forsaken?599(Xu et al., 2012) Modern have you given up so quickly on rosaline, whom you loved so much?
JFLEG Corpus Ungrammatical Forexample, My cousin is 12years old.747(Napoles et al., 2017) Grammatical For example, my cousin is 12 years old.
Symbolic Manipulation Symbolic olive >cat1000(Ours ) English olive is greater than cat
Table 2: Overview of the textual style transfer datasets used in this paper.
5 Experiments and Results
5.1 Datasets
Differently from most previous work, which focused on single TST subtasks or datasets, we
present experiments on a wide range of TST subtasks (also described in Table 2):
•YELP: Sentiment transfer for Yelp reviews
(Zhang et al., 2015).
•AMAZON : Sentiment transfer for Amazon reviews (Li et al., 2018).
•SHAKESPEARE : Elizabethan -to-modern
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
27
|
•AMAZON : Sentiment transfer for Amazon reviews (Li et al., 2018).
•SHAKESPEARE : Elizabethan -to-modern
translation for Shakespeare (Xu et al., 2012).
•GYAFC : Formality transfer for Yahoo Answers responses (Li et al., 2018).
•JFLEG : Grammar error correction for student
essays (Napoles et al., 2017).
•SYM: Symbol-to-natural-language translation
on a new custom synthetic dataset.
In the initial stages of our research, we noticed
that all of these datasets, with the exception of
SYM(which is synthetic), contain various tokenization issues (e.g., sentences sometimes contain extra white-space or have their punctuation marks
separated out by spaces). We did not wish these
tokenization artifacts to diminish the quality of our
generations from general-purpose LMs—neither
did we want this issue to negatively impact our
evaluation scheme. To that end, we used a simple
text-cleaning procedure to clean the texts.16
5.2 Evaluation Metrics
Prior studies on style and sentiment transfer have
typically evaluated models across three dimensions:
content/meaning preservation (textual similarity),
style transfer strength, and fluency (Mir et al., 2019;
16We release both the original and cleaned versions of the
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
28
|
content/meaning preservation (textual similarity),
style transfer strength, and fluency (Mir et al., 2019;
16We release both the original and cleaned versions of the
datasets alongside this paper to help facilitate future research.
In the Appendix, we also present results for both the original
and cleaned datasets.Briakou et al., 2021). We note that remarkably,
these dimensions correspond exactly to the criteria
that appear in Equation 1 in §3.1.
Content Preservation. BLEU (Papineni et al.,
2002) is the standard metric for measuring semantic content preservation. We use the SacreBLEU
(sBLEU) implementation (Post, 2018) to compute
both reference -BLEU ( r-sBLEU) and self-sBLEU
(s-sBLEU) scores. Whereas r-sBLEU helps measure the distance of generated sentences from the
ground-truth references, s-sBLEU indicates the degree to which the model directly copies the source.
Transfer Strength. In order to determine
whether outputs generated by a TST model have
the attributes of their target styles, we follow the
standard classifier-based approach: we train a (binary) style classifier on the corpus of interest and
use it to estimate the fraction of generated outputs
whose styles match their target styles.
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
29
|
use it to estimate the fraction of generated outputs
whose styles match their target styles.
Fluency. To measure the fluency of generated
texts, we compute their average token-level perplexity (PPL) using a pre-trained LM (in our case,
GPT-2-Large). We note that, whilst this PPL-driven
approach has the advantage of being automated and
practical, it still contains considerable drawbacks,
including biases towards shorter texts.
5.3 Model Choices.
We used four GPT-2 models (Radford et al., 2019)
of varying sizes (viz., GPT-2-Small (117M params),
GPT-2-Medium (345M), GPT-2-Large (774M),
and GPT-2-XL (1.6B)), GPT-Neo-1.3B Black et al.
(2021), GPT-Neo-2.7B, and GPT-J-6B (Wang and
Komatsuzaki, 2021). We highlight that none of
these models were fine-tuned or prompt-tuned.
5.4 Results
Here, we present a summary of our key findings.
For our complete results, we encourage the reader
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
30
|
5.4 Results
Here, we present a summary of our key findings.
For our complete results, we encourage the reader
to see the Appendix (especially, Tables 10-19).
6
Model Acc r-sBLEU s-sBLEU PPL
SUPERVISED
[1] CrossAlignment 0.73 7.8 18.3 217
[2] BackTrans 0.95 2.0 46.5 158
[3] MultiDecoder 0.46 13.0 39.4 373
[4] DeleteOnly 0.85 13.4 33.9 182
[4] DeleteAndRetrieve 0.90 14.7 36.4 180
[5] UnpairedRL 0.49 16.8 45.7 385
[6] DualRL 0.88 25.9 58.9 133
[7] ST (Multi-Class) 0.86 26.4 63.0 175
[7] ST (Conditional) 0.93 22.9 52.8 223
[8] B-GST 0.81 21.6 46.5 158
ZERO-ORFEW-SHOT INFERENCE ONLY
[9] LLM Aug-0S-FirstChoice 0.85 5.3 9.2 33
[9] LLM 5S-FirstChoice 0.93 6.7 11.2 43
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
31
|
[9] LLM 5S-FirstChoice 0.93 6.7 11.2 43
[9] LLM Aug-0S-Best-sBLEU†0.63 19.8 45.1 55
[9] LLM 5S-Best-sBLEU†0.78 23.2 48.3 77
Ours (GPT-2-XL) 0.87 14.8 28.7 65
Ours (GPT-J-6B) 0.87 23.0 47.7 80
Table 3: A comparison of our Prompt-and-Rerank approach
with supervised sentiment transfer methods and the ultralarge-scale prompting-based method of Reif et al. (2022) on
theYELP-clean dataset. In order to compare fairly against
previous studies, we applied our data-cleaning code to their
publicly-available outputs and re-computed all evaluation metrics. References: [1] (Shen et al., 2017), [2] (Prabhumoye
et al., 2018), [3] (Fu et al., 2018), [4] (Li et al., 2018), [5] (Xu
et al., 2018), [6] (Luo et al., 2019b), [7] (Dai et al., 2019),
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
32
|
et al., 2018), [6] (Luo et al., 2019b), [7] (Dai et al., 2019),
[8] (Sudhakar et al., 2019), [9] (Reif et al., 2022). Note on†:
We used sBLEU to choose the best candidate, as opposed to
BLEU that was used originally in (Reif et al., 2022).
Dataset Model Accr-sBLEU s-sBLEU PPL
AMAZON GPT-2-XL 0.70 11.5 17.2 77
P!N GPT-J-6B 0.65 21.5 31.4 70
AMAZON GPT-2-XL 0.56 13.2 19.9 50
N!P GPT-J-6B 0.52 19.3 29.3 58
YELP GPT-2-XL 0.87 14.8 28.7 65
P!N GPT-J-6B 0.87 23.0 47.7 80
YELP GPT-2-XL 0.72 12.0 25.3 55
N!P GPT-J-6B 0.65 20.2 44.6 58
SHAKE - GPT-2-XL 0.39 18.9 38.4 90
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
33
|
N!P GPT-J-6B 0.65 20.2 44.6 58
SHAKE - GPT-2-XL 0.39 18.9 38.4 90
SPEARE GPT-J-6B 0.78 21.9 31.8 81
JFLEGGPT-2-XL 35.9 74.8 91.5 76
GPT-J-6B 40.0 64.8 59.1 48
GYAFCGPT-2-XL 0.82 32.7 41.9 58
GPT-Neo-1.3B 0.85 36.4 49.6 68
SYMGPT-2-XL 0.56 68.5 - GPT-J-6B 0.74 81.9 - Table 4: Four-shot performances of GPT-2-XL and GPT-J
across all style transfer tasks, using curly brackets as delimiters. Full results with all models and delimiter pairs are shown
in the appendix. P!Nstands for the positive !negative direction, and vice-versa for N!P.for JFLEG, GLEU score
is used in place of accuracy to measure grammar correction
performance. Note that the r-sBLEU column is not bolded because it is not necessarily desirable to have a higher r-sBLEU.
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
34
|
performance. Note that the r-sBLEU column is not bolded because it is not necessarily desirable to have a higher r-sBLEU.
Table 3 juxtaposes our results on YELP with
those of prior studies. Despite not training or finetuning, our method is competitive with prior models that were designed and trained specifically for
these tasks. In fact, compared to supervised methods, our models almost always generate more fluent outputs, as measured by perplexity. Compared
to Reif et al. (2022), who utilize the proprietary
137-billion-parameter LLM (LaMDA), we compareSetting Acc r-sBLEU s-sBLEU PPL
Vanilla 78.0 14.7 31.0 58.5
Contrastive 79.5 13.4 27.0 59.5
Negation-v1 66.5 13.4 28.1 67.5
Negation-v2 52.0 18.0 40.6 69.0
Table 5: Four-shot performances of GPT-2-XL on the YELPclean dataset under different prompting protocols. We show
the average of scores from P!NandN!Pdirections. A full
table with all models is included in the Appendix. Across all
models, the vanilla and contrastive prompting protocols tend
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
35
|
table with all models is included in the Appendix. Across all
models, the vanilla and contrastive prompting protocols tend
to yield the most favourable results.
Delim. Acc r-sBLEU s-sBLEU PPL
hi 49.5 17.4 40.8 45
* "" 55.0 12.0 29.8 37
i"" 53.0 10.7 25.4 35
fg 59.5 10.0 23.6 35
–– 54.5 6.4 16.5 24
ffgg 50.5 18.3 43.9 65
() 55.5 12.4 28.1 43
"" 60.5 8.8 20.4 31
[] 58.0 11.4 27.4 41
Table 6: Zero-shot performances of GPT-2-XL on the YELPclean dataset using different delimiter pairs. Full tables with
all models for all datasets are included in the Appendix.
on-par or favorably despite using much smaller
models; we obtain better sBLEU scores than their
“FirstChoice” setting (which uses a single output)
and better accuracy scores than their “BestBLEU”
oracle setting (which takes the best of 16 outputs,
as measured by sBLEU score).
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
36
|
and better accuracy scores than their “BestBLEU”
oracle setting (which takes the best of 16 outputs,
as measured by sBLEU score).
Table 4 presents a summary of our results across
all seven TST datasets for GPT-2-XL and GPT-J.
For full results including all models (GPT-2-Small
to GPT-J), please refer to the Appendix. Broadly,
we find that all models are capable of TST to a
reasonable degree—with the larger models (e.g.,
GPT-2-XL, GPT-Neo-2.7B, GPT-J) often performing better than the smaller models. The only model
that consistently performs poorly is GPT-2-Small:
Its high s-sBLEU and low accuracy indicate that it
copies long sections of the input (without changing
its style) more often than the other models.
Looking at individual tasks, we recognize that
there remains substantial room for improvement on
theJFLEG task: Most models underperformed a
simple baseline that copied the input text without
making any changes. The baseline achieved 37.2
GLEU, better than all models except GPT-J (which
obtained 40.0). Finally, on our new synthetic task
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
37
|
making any changes. The baseline achieved 37.2
GLEU, better than all models except GPT-J (which
obtained 40.0). Finally, on our new synthetic task
SYM, we found that GPT-J performed significantly
better than the rest: It achieved 74% accuracy17
whereas no other model exceeded 60% accuracy.18
17Accuracy is measured via exact-string-matching.
18When the models failed to generate the correct output,
we found that a common failure case was copying the input
words correctly but using the wrong logic (e.g., generating
“less than” instead of “greater than”).
7
5.5 Further Analysis and Discussion
Contrastive prompting generally improves
style transfer quality. As shown in Table 5
(and Table 19 in the Appendix), amongst the four
prompting protocols considered in this paper, contrastive prompting generally yielded the best accuracy, albeit not always the best sBLEU scores.
Delimiter-pair choice has a large impact on
model performance. Our systematic analysis of
ten different delimiter-pairs shows that delimiter
choice substantially affects the quality of generated
outputs. Although there is not a single pair which
performs best across all setups, certain delimiters,
such as the curly brackets fg, square brackets [],
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
38
|
outputs. Although there is not a single pair which
performs best across all setups, certain delimiters,
such as the curly brackets fg, square brackets [],
parentheses (), and quotes "", yielded consistently better results on both AMAZON andYELP
(see Tables 9-12). We hypothesize that the strong
performance of these markers is attributable to the
fact that they are often used as text separators (or
dividers) in different textual contexts, such as essays, dialogues, and code snippets, which compose
part of the pre-training data of our models.
Re-ranking method improves overall performance. We considered two re-ranking approaches,
one in which we picked the generated output with
the highest beam score and one in which we sampled three outputs from the model using beam
search and then re-scored them according to three
criteria discussed in §3.2. As shown in Tables 14
and 15, the re-ranking method can boost the sentiment accuracy by 10-30%. It often, but not always,
leads to better sBLEU and fluency scores. Also, as
Table 7 illustrates, if we have access to a classifier
trained on paired data, it might be more convenient
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
39
|
Table 7 illustrates, if we have access to a classifier
trained on paired data, it might be more convenient
to use it in our style transfer accuracy measurements, instead of an MLM as a proxy-classifier, in
the re-ranking process, as it empirically leads to
higher accuracy and sBLEU scores.
Analysis of bias and transfer performance in
opposite directions. We find that pre-trained models have strong directional biases: None of the
models performed the same when going in the
negative!positive ( N!P) and positive!negative
(P!N) directions on AMAZON andYELP. We
offer three possible explanations for this phenomenon: (i) The inherent differences in the linguistic difficulty of the tasks, (ii) the potential biases in pre-training dataset(s), and (iii) the poor
quality of annotations in certain style transfer directions. Regarding (i), a qualitative inspection of the
sentiment transfer datasets illustrates that in someModel Setting Acc r-sBLEU s-sBLEU PPL
GPT-2-XL
(1558M)Top Choice 0.63 13.7 20.3 65
P&R RoBERTa 0.87 14.8 28.7 65
P&R Oracle Cl. 0.95 16.8 33.4 63
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
40
|
P&R RoBERTa 0.87 14.8 28.7 65
P&R Oracle Cl. 0.95 16.8 33.4 63
GPT-J-6B
(6B)Top Choice 0.81 25.3 50.5 107
P&R RoBERTa 0.87 23.0 47.7 80
P&R Oracle Cl. 0.95 25.4 52.4 87
Table 7: Comparison of vanilla four-shot performance of
GPT-2 XL and GPT-J-6B models on YELP-clean (P!N)
under three settings: (1) choosing the output with the highest
beam score (TC), (2) Prompt-and-Rerank with RoBERTa used
as a zero-shot style classifier (P&R RoBERTa ), and (3) Promptand-Rerank with an oracle style classifier trained on paired
data (P&R Oracle). For full results in YELP-clean, see Table 15.
cases, good P!Nperformance can be achieved
by simply adding a negation (e.g., “not”) into the
text. Regarding (ii), it is possible that the webscraped pre-training data of these models contains
more sentences that resemble the task of changing
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
41
|
text. Regarding (ii), it is possible that the webscraped pre-training data of these models contains
more sentences that resemble the task of changing
the sentiment from positive to negative than the
reverse direction during their pre-training periods.
Qualitatively, the GPT-2 models appear adept at
negation; therefore, it may not be surprising that
these models yield better results in the P!Ndirection. As for (iii), our inspection of the ground-truth
data reveals that it contains some noisy labels and
incorrect input-output pairs.
Limitations. The primary limitation of our reranking method is that it involves generating multiple outputs from an autoregressive LM, which
requires multiple forward passes. Additionally, our
approach relies on having access to a pre-trained
bi-directional MLM. Compared to a simple zeroshot approach, these elements could potentially add
complexity to deploying this model in practice.
6 Conclusion
In this paper, we propose a novel formal framework for textual style transfer. This framework naturally leads us to a new method, which we denote
Prompt-and-Rerank , that utilizes general-purpose
pretrained language models to transform text into
in arbitrary styles. In our experiments, we use
our method to demonstrate that off-the-shelf, pretrained “small” language models, such as GPT-2,
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
42
|
in arbitrary styles. In our experiments, we use
our method to demonstrate that off-the-shelf, pretrained “small” language models, such as GPT-2,
can perform arbitrary textual style transfer, without
any additional model fine-tuning or prompt-tuning.
Additionally, we conduct an extensive investigation
prompt phrasing and delimiter choice on transfer
quality. In total, we hope that our work makes further research in this area more accessible to a broad
set of researchers, both by alleviating the computational constraints of hundred-billion-parameter
language models and by establishing a standard set
of clean datasets for arbitrary text style transfer.
8
7 Acknowledgments
We would like to thank Dora Demszky, Esin Durmus, Tatsu Hashimoto, John Hewitt, Faisal Ladhak,
Percy Liang, Nelson Liu, Shikhar Murty, Tolúl o.pé.
Ògúnr è.mí, Isabel Papadimitriou, Ashwin Paranjape, Suproteem K. Sarkar, Stuart M. Shieber, Kyle
Swanson, Rose Wang, and Tianyi Zhang for their
help and support, constructive comments, and valuable suggestions. We also thank Sudha Rao for
help with the navigation of the GYAFC dataset
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
43
|
help and support, constructive comments, and valuable suggestions. We also thank Sudha Rao for
help with the navigation of the GYAFC dataset
and Yunli Wang for sharing their data splits and
classifiers for GYAFC that were used in their paper.
Melas-Kyriazi gratefully acknowledges the support
of the Rhodes Trust.
References
Sid Black, Leo Gao, Phil Wang, Connor Leahy, and
Stella Biderman. 2021. GPT-Neo: Large Scale
Autoregressive Language Modeling with MeshTensorflow.
Eleftheria Briakou, Sweta Agrawal, Joel Tetreault,
and Marine Carpuat. 2021. Evaluating the Evaluation Metrics for Style Transfer: A Case Study
in Multilingual Formality Transfer. arXiv preprint
arXiv:2110.10668 .
Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie
Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind
Neelakantan, Pranav Shyam, Girish Sastry, Amanda
Askell, Sandhini Agarwal, Ariel Herbert-V oss,
Gretchen Krueger, Tom Henighan, Rewon Child,
Aditya Ramesh, Daniel M. Ziegler, Jeffrey Wu,
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
44
|
Gretchen Krueger, Tom Henighan, Rewon Child,
Aditya Ramesh, Daniel M. Ziegler, Jeffrey Wu,
Clemens Winter, Christopher Hesse, Mark Chen,
Eric Sigler, Mateusz Litwin, Scott Gray, Benjamin
Chess, Jack Clark, Christopher Berner, Sam McCandlish, Alec Radford, Ilya Sutskever, and Dario
Amodei. 2020. Language Models are Few-Shot
Learners. In Advances in Neural Information Processing Systems 33: Annual Conference on Neural Information Processing Systems 2020, NeurIPS
2020, December 6-12, 2020, virtual .
Ning Dai, Jianze Liang, Xipeng Qiu, and Xuan-Jing
Huang. 2019. Style Transformer: Unpaired Text
Style Transfer without Disentangled Latent Representation. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics ,
pages 5997–6007.
Jacob Devlin, Ming-Wei Chang, Kenton Lee, and
Kristina Toutanova. 2019. BERT: Pre-training of
Deep Bidirectional Transformers for Language Understanding. In Proceedings of the 2019 Conference
of the North American Chapter of the Association
for Computational Linguistics: Human Language
Technologies, Volume 1 (Long and Short Papers) ,
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
45
|
of the North American Chapter of the Association
for Computational Linguistics: Human Language
Technologies, Volume 1 (Long and Short Papers) ,
pages 4171–4186, Minneapolis, Minnesota. Association for Computational Linguistics.Zhenxin Fu, Xiaoye Tan, Nanyun Peng, Dongyan Zhao,
and Rui Yan. 2018. Style Transfer in Text: Exploration and Evaluation. In Proceedings of the AAAI
Conference on Artificial Intelligence , volume 32.
Tianyu Gao, Adam Fisch, and Danqi Chen. 2021. Making Pre-trained Language Models Better Few-shot
Learners. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics
and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers) ,
pages 3816–3830.
Ruining He and Julian McAuley. 2016. Ups and
Downs: Modeling the Visual Evolution of Fashion
Trends with One-Class Collaborative Filtering. In
proceedings of the 25th international conference on
world wide web , pages 507–517.
Zhiting Hu, Zichao Yang, Xiaodan Liang, Ruslan
Salakhutdinov, and Eric P Xing. 2017. Toward
Controlled Generation of Text. In International
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
46
|
Salakhutdinov, and Eric P Xing. 2017. Toward
Controlled Generation of Text. In International
conference on machine learning , pages 1587–1596.
PMLR.
Di Jin, Zhijing Jin, Zhiting Hu, Olga Vechtomova,
and Rada Mihalcea. 2021. Deep Learning for Text
Style Transfer: A Survey. Computational Linguistics, pages 1–51.
Kalpesh Krishna, John Wieting, and Mohit Iyyer. 2020.
Reformulating Unsupervised Style Transfer as Paraphrase Generation. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language
Processing (EMNLP) , pages 737–762, Online. Association for Computational Linguistics.
Guillaume Lample, Sandeep Subramanian, Eric Smith,
Ludovic Denoyer, Marc’Aurelio Ranzato, and YLan Boureau. 2019. Multiple-Attribute Text Rewriting. In International Conference on Learning Representations .
Juncen Li, Robin Jia, He He, and Percy Liang. 2018.
Delete, Retrieve, Generate: a Simple Approach to
Sentiment and Style Transfer. In Proceedings of the
2018 Conference of the North American Chapter of
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
47
|
Delete, Retrieve, Generate: a Simple Approach to
Sentiment and Style Transfer. In Proceedings of the
2018 Conference of the North American Chapter of
the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers) , pages 1865–1874, New Orleans, Louisiana.
Association for Computational Linguistics.
Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis,
Luke Zettlemoyer, and Veselin Stoyanov. 2019.
RoBERTa: A Robustly Optimized BERT Pretraining Approach. CoRR , abs/1907.11692.
Fuli Luo, Peng Li, Pengcheng Yang, Jie Zhou, Yutong Tan, Baobao Chang, Zhifang Sui, and Xu Sun.
2019a. Towards Fine-Grained Text Sentiment Transfer. In Proceedings of the 57th Annual Meeting
of the Association for Computational Linguistics ,
pages 2013–2022.
9
Fuli Luo, Peng Li, Jie Zhou, Pengcheng Yang, Baobao
Chang, Zhifang Sui, and Xu Sun. 2019b. A
Dual Reinforcement Learning Framework for Unsupervised Text Style Transfer. arXiv preprint
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
48
|
Chang, Zhifang Sui, and Xu Sun. 2019b. A
Dual Reinforcement Learning Framework for Unsupervised Text Style Transfer. arXiv preprint
arXiv:1905.10060 .
Aman Madaan, Amrith Setlur, Tanmay Parekh, Barnabás Póczos, Graham Neubig, Yiming Yang, Ruslan
Salakhutdinov, Alan W Black, and Shrimai Prabhumoye. 2020. Politeness Transfer: A Tag and Generate Approach. In Proceedings of the 58th Annual
Meeting of the Association for Computational Linguistics , pages 1869–1881.
Eric Malmi, Aliaksei Severyn, and Sascha Rothe.
2020. Unsupervised Text Style Transfer with
Padded Masked Language Models. arXiv preprint
arXiv:2010.01054 .
Remi Mir, Bjarke Felbo, Nick Obradovich, and Iyad
Rahwan. 2019. Evaluating Style Transfer for Text.
arXiv preprint arXiv:1904.02295 .
Courtney Napoles, Keisuke Sakaguchi, Matt Post, and
Joel Tetreault. 2015. Ground Truth for Grammatical Error Correction Metrics. In Proceedings of the
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
49
|
Courtney Napoles, Keisuke Sakaguchi, Matt Post, and
Joel Tetreault. 2015. Ground Truth for Grammatical Error Correction Metrics. In Proceedings of the
53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint
Conference on Natural Language Processing (Volume 2: Short Papers) , pages 588–593.
Courtney Napoles, Keisuke Sakaguchi, and Joel
Tetreault. 2017. JFLEG: A fluency corpus and
benchmark for grammatical error correction. In Proceedings of the 15th Conference of the European
Chapter of the Association for Computational Linguistics: Volume 2, Short Papers , pages 229–234,
Valencia, Spain. Association for Computational Linguistics.
Kishore Papineni, Salim Roukos, Todd Ward, and WeiJing Zhu. 2002. BLEU: A method for Automatic
Evaluation of Machine Translation. In Proceedings
of the 40th annual meeting of the Association for
Computational Linguistics , pages 311–318.
Fabio Petroni, Tim Rocktäschel, Sebastian Riedel,
Patrick Lewis, Anton Bakhtin, Yuxiang Wu, and
Alexander Miller. 2019. Language Models as
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
50
|
Patrick Lewis, Anton Bakhtin, Yuxiang Wu, and
Alexander Miller. 2019. Language Models as
Knowledge Bases? In Proceedings of the 2019 Conference on Empirical Methods in Natural Language
Processing and the 9th International Joint Conference on Natural Language Processing (EMNLPIJCNLP) , pages 2463–2473.
Matt Post. 2018. A Call for Clarity in Reporting BLEU
Scores. In Proceedings of the Third Conference on
Machine Translation: Research Papers , pages 186–
191, Brussels, Belgium. Association for Computational Linguistics.
Shrimai Prabhumoye, Rafal Kocielnik, Mohammad
Shoeybi, Anima Anandkumar, and Bryan Catanzaro.
2021. Few-shot Instruction Prompts for Pretrained
Language Models to Detect Social Biases. arXiv
preprint arXiv:2112.07868 .Shrimai Prabhumoye, Yulia Tsvetkov, Ruslan Salakhutdinov, and Alan W Black. 2018. Style Transfer
Through Back-Translation. In Proceedings of the
56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages
866–876.
Alec Radford, Jeffrey Wu, Rewon Child, David Luan,
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
51
|
866–876.
Alec Radford, Jeffrey Wu, Rewon Child, David Luan,
Dario Amodei, Ilya Sutskever, et al. 2019. Language Models are Unsupervised Multitask Learners.
OpenAI blog , 1(8):9.
Sudha Rao and Joel Tetreault. 2018. Dear Sir or
Madam, May I Introduce the GYAFC Dataset: Corpus, Benchmarks and Metrics for Formality Style
Transfer. In Proceedings of the 2018 Conference of
the North American Chapter of the Association for
Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers) , pages 129–140,
New Orleans, Louisiana. Association for Computational Linguistics.
Machel Reid and Victor Zhong. 2021. LEWIS: Levenshtein Editing for Unsupervised Text Style Transfer.
arXiv preprint arXiv:2105.08206 .
Emily Reif, Daphne Ippolito, Ann Yuan, Andy Coenen, Chris Callison-Burch, and Jason Wei. 2022. A
Recipe for Arbitrary Text Style Transfer with Large
Language Models. In Proceedings of the 60th Annual Meeting of the Association for Computational
Linguistics (Volume 2: Short Papers) , pages 837–
848, Dublin, Ireland. Association for Computational
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
52
|
Linguistics (Volume 2: Short Papers) , pages 837–
848, Dublin, Ireland. Association for Computational
Linguistics.
Tianxiao Shen, Tao Lei, Regina Barzilay, and Tommi
Jaakkola. 2017. Style Transfer from Non-Parallel
Text by Cross-Alignment. Advances in neural information processing systems , 30.
Akhilesh Sudhakar, Bhargav Upadhyay, and Arjun Maheswaran. 2019. "Transforming" Delete, Retrieve,
Generate Approach for Controlled Text Style Transfer. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing
and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP) , pages
3269–3279.
Ben Wang and Aran Komatsuzaki. 2021. GPTJ-6B: A 6 Billion Parameter Autoregressive
Language Model. https://github.com/
kingoflolz/mesh-transformer-jax .
Yunli Wang, Yu Wu, Lili Mou, Zhoujun Li, and Wenhan Chao. 2019. Harnessing Pre-Trained Neural
Networks with Rules for Formality Style Transfer.
InProceedings of the 2019 Conference on Empirical
Methods in Natural Language Processing and the
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
53
|
Networks with Rules for Formality Style Transfer.
InProceedings of the 2019 Conference on Empirical
Methods in Natural Language Processing and the
9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP) , pages 3573–
3578, Hong Kong, China. Association for Computational Linguistics.
10
Yunli Wang, Yu Wu, Lili Mou, Zhoujun Li, and Wenhan Chao. 2020. Formality Style Transfer with
Shared Latent Space. In Proceedings of the 28th
International Conference on Computational Linguistics, pages 2236–2249.
Xing Wu, Tao Zhang, Liangjun Zang, Jizhong Han,
and Songlin Hu. 2019. "Mask and Infill": Applying Masked Language Model to Sentiment Transfer.
arXiv preprint arXiv:1908.08039 .
Jingjing Xu, Xu Sun, Qi Zeng, Xiaodong Zhang, Xuancheng Ren, Houfeng Wang, and Wenjie Li. 2018.
Unpaired Sentiment-to-Sentiment Translation: A
Cycled Reinforcement Learning Approach. In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long
Papers) , pages 979–988.
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
54
|
Papers) , pages 979–988.
Wei Xu, Alan Ritter, Bill Dolan, Ralph Grishman, and
Colin Cherry. 2012. Paraphrasing for Style. In COLING, pages 2899–2914.
Zhilin Yang, Zihang Dai, Yiming Yang, Jaime Carbonell, Russ R Salakhutdinov, and Quoc V Le. 2019.
XLNet: Generalized Autoregressive Pretraining for
Language Understanding. Advances in neural information processing systems , 32.
Tianyi Zhang, Varsha Kishore, Felix Wu, Kilian Q.
Weinberger, and Yoav Artzi. 2020. BERTScore:
Evaluating Text Generation with BERT. In International Conference on Learning Representations .
Xiang Zhang, Junbo Zhao, and Yann LeCun. 2015.
Character-Level Convolutional Networks for Text
Classification. Advances in neural information processing systems , 28:649–657.
11
A Appendix
A.1 Additional Details about Datasets
Previous TST studies have often chosen to focus on
particular subtasks (such as changing the sentiment
of a text from positive to negative) or particular
datasets (such as YELPorAMAZON ). In contrast,
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
55
|
particular subtasks (such as changing the sentiment
of a text from positive to negative) or particular
datasets (such as YELPorAMAZON ). In contrast,
in our experiments, we decided to focus on a variety of TST datasets, some of which are known
and widely used datasets in the field and some of
which are new and synthetic. In the first half of this
section, we present and discuss these datasets.19
Yelp Sentiment Dataset. YELPis a subset of the
Yelp Review Polarity Dataset that was first used by
Zhang et al. (2015) for a text classification task. It
consists restaurant and other business reviews from
Yelp, along with a label—either positive ornegative—for each review. We used the version of the
dataset that was curated by Li et al. (2018) in our
experiments. The test set contains 500 positive and
500 negative samples, with one human reference
(ground-truth) for each sample.
Amazon Sentiment Dataset. AMAZON is similar to YELPin its nature, but it contains product
reviews that were obtained from Amazon. Each
review is labeled either positive ornegative . As
before, we used the version of the dataset that was
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
56
|
reviews that were obtained from Amazon. Each
review is labeled either positive ornegative . As
before, we used the version of the dataset that was
used by Li et al. (2018). The test set contains 500
positive and 500 negative sentences, with one human reference output for each sample.
Shakespearean English Dataset. We additionally used a small subset of the dataset that was used
by Xu et al. (2012) originally for phrase-based machine translation, and experimented with “translating” sentences written in Elizabethan English
tomodern English. This small test set, which we
callSHAKESPEARE , contains 599 paired sentences
from William Shakespeare’s Romeo and Juliet ,
written in Elizabethan and modern English.20
GYAFC Formality Dataset. Grammarly’s Yahoo Answers Formality corpus ( GYAFC ; Rao and
19We chose these datasets to broaden the semantic diversity
of the TST tasks and to establish benchmarks for new TST
studies. We share both the original and clean versions of
some of the widely-used but poorly-tokenized datasets, such
asAMAZON andYELP. In doing so, we hope to help address
the recent call-to-action on reproducibility in TST from Jin
et al. (2021); they encouraged researchers to share their data
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
57
|
the recent call-to-action on reproducibility in TST from Jin
et al. (2021); they encouraged researchers to share their data
and evaluation codes in order to establish reliable benchmarks
and facilitate easier comparison of new studies with existing
work. We hope that our efforts will be a constructive step
towards this goal.
20All the input sentences in SHAKESPEARE contain at least
10 and at most 25 words (inclusive).Tetreault (2018)) contains paired informal andformalsentences. Following Luo et al. (2019b), we
used the samples from the “Family & Relationship”
(F&R) domain and restricted our focus to the informal to formal direction. The test set contains 500
formal and 500 informal sentences.
JFLEG Corpus. The JHU FLuency-Extended
GUG ( JFLEG ) Corpus was introduced by Napoles
et al. (2017) to train and evaluate models for automatic grammatical error correction. It contains
paired grammatical andungrammatical sentences
(with three error types—namely, awkward, orthographic, and grammatical). In our experiments, we
focused on the ungrammatical to grammatical direction and used the publicly available test set that
contains 747 sentences.
Symbolic Manipulation Task. We designed
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
58
|
focused on the ungrammatical to grammatical direction and used the publicly available test set that
contains 747 sentences.
Symbolic Manipulation Task. We designed
this small synthetic dataset to investigate how skillful the off-the-shelf SLMs are at writing symbolic
expressions as natural English-language sentences.
This dataset contains 1,000 example pairs, in which
each input sample is written in a symbolic form (as
either “ > ” or “ < ”, where and are
two different single words from the animal color,
fruit, and number categories) and its corresponding
output is basically the spoken utterance in English.
Remark. We realized that the original versions of
all the aforementioned real-world TST datasets contain various tokenization issues (for instance, sentences sometimes contain extra whitespaces or have
their puntuation marks separated out by spaces).
We did not wish these tokenization artifacts to diminish the quality of our generations. To that end,
we used a simple text-cleaning procedure to clean
the texts before feeding them to our models.21
A.2 Additional Evaluation Metrics
Here, we describe in greater detail the standard automatic evaluation metrics used in the assessment
of the performance of TST models.
Content Preservation. The standard metric for
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
59
|
Here, we describe in greater detail the standard automatic evaluation metrics used in the assessment
of the performance of TST models.
Content Preservation. The standard metric for
measuring semantic content preservation (or textual similarity, as we call it) has been BLEU (Papineni et al., 2002): If reference (ground-truth) sentences are available, then reference -BLEU scores
are calculated by comparing model outputs to
human-written ground-truth outputs using n-grams.
Some recent studies (Lample et al., 2019; Dai et al.,
2019) further look at self-BLEU scores, compar21For the AMAZON andYELPdataset, we show the benefit
of data-cleaning on overall performance. We also publicly
release our text-cleaning code.
12
ing model outputs to input sentences—this is particularly done when reference sentences are not
directly available. In our evaluations, we primarily used the SacreBLEU metric (Post, 2018)—as
SacreBLEU has been shown to be a more reliable
and accessible metric than BLEU—and considered
both reference -SacreBLEU ( r-sBLEU) and selfSacreBLEU ( s-sBLEU) scores.22When evaluating
the performances of models on the JFLEG corpus, we also used the sentence-level GLEU metric
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
60
|
the performances of models on the JFLEG corpus, we also used the sentence-level GLEU metric
(Napoles et al., 2015), a variant of BLEU that was
specifically designed for evaluating grammatical
error correction (GEC) models.
Transfer Strength. To determine whether outputs generated by a TST model have the attributes
of their target styles, the most common approach
has been to train a (binary) classifier on the training
set of the corpus of focus, where the sentences are
taken as the inputs and their corresponding styles
as the labels, and then to use this trained classifier
to predict the percentage of the generated outputs
for which the styles predicted by the model match
their target styles.23In our sentiment transfer experiments, we measured transfer strength (sentiment
accuracy) by fine-tuning pre-trained RoBERTa classifiers (Liu et al., 2019) on the training data in each
case. In our experiments on SHAKESPEARE , we
used the RoBERTa-based Shakespeare classifier of
Krishna et al. (2020). Finally, in our experiments
onGYAFC , we fine-tuned a pre-trained RoBERTa
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
61
|
Krishna et al. (2020). Finally, in our experiments
onGYAFC , we fine-tuned a pre-trained RoBERTa
classifier on a subset of F&R examples.24
Fluency. With the emergence of successful LMs
at our disposal, most recent TST models measure
the fluency of their generated texts by computing
perplexity (PPL) via a pre-trained LM like GPT2.25Whilst this PPL-driven approach has the advantage of being automated and practical, it still
contains considerable drawbacks, among which biases towards short texts and more frequent tokens
can be listed right away. In our evaluations, we
reported the average token-level PPL of generated
texts using GPT-2-Large (774M).
22We used the SacreBLEU metric implemented in Hugging Face’s Metrics library and lowered all the texts—both
predictions and references—before calculating the scores.
23This method of measuring transfer accuracy demands
access to either paired data for training a classifier or a pretrained classier that can accurately estimate the style of an
input text. It is, therefore, difficult to measure transfer accuracy for arbitrary or unknown styles, because there may not
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
62
|
input text. It is, therefore, difficult to measure transfer accuracy for arbitrary or unknown styles, because there may not
be any specific data to train a classifier.
24We release all our fine-tuned classifiers on our codebase.
25Early work used to measure fluency of sentences using
ann-gram (typically trigram) Kneser-Ney language model.A.3 Full Results
In the tables below, we include zero-shot results
for the clean versions of AMAZON (Table 10) and
YELP(Table 12), as well as the original versions
ofAMAZON (Table 9) and YELP(Table 11). We
also include four-shot results for the clean versions
ofAMAZON (Table 13), YELP(Table 14), SHAKE SPEARE (Table 15), JFLEG (Table 16), GYAFC
(Table 17), and S YM(Table 18).
A.4 Further Discussion
Sentiment Transfer. Table 14 and Table 15 show
the results for the clean versions of AMAZON and
YELP, respectively. In terms of sentiment accuracy, GPT-2-XL yielded the best performance on
both datasets, achieving 70% (87%) positive !
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
63
|
YELP, respectively. In terms of sentiment accuracy, GPT-2-XL yielded the best performance on
both datasets, achieving 70% (87%) positive !
negative accuracy and 56% (72%) negative !positive on AMAZON (YELP). In both cases, however,
the sBLEU scores of GPT-2-XL were relatively
lower than those of other models, indicating that it
copied more from the source text. The GPT-Neo
models had higher r-sBLEU and s-sBLEU scores
than GPT-2-XL on both AMAZON andYELP, with
only slightly worse accuracy scores. In the case
ofYELP-clean especially, the GPT-Neo/J models
achieved good balances of sentiment accuracy, textual similarity, and fluency.
Shakespeare-to-Modern English Translation.
As shown in Table 13, model performance generally improves with model size, with GPT-J-6B
achieving almost 80% accuracy (according to the
supervised classifier) and 21.9 r-sBLEU. Also notable is the difference between GPT-2-Small’s high
s-sBLEU score and low classifier accuracy, relative
to the other models. Together, these indicate that
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
64
|
s-sBLEU score and low classifier accuracy, relative
to the other models. Together, these indicate that
the model copies large parts of the input text more
often than the other GPT models.
Formality Transfer and Grammatical Error
Correction . For GYAFC (Table 17), most models
achieved accuracy scores above 80%, with increasing model size correlating with BLEU score. Notably, GPT-Neo-2.7B achieved an accuracy score
of81% and and a r-sBLEU score of 50 in the informal to formal direction. For JFLEG (Table 16), on
the other hand, most models failed to outperform a
simple baseline, which automatically copied the input text without making any changes. This baseline
achieves a GLEU score of 37.2, better than all models except GPT-J (which obtains 40.0). Broadly,
there remains substantial room for improvement on
JFLEG.
13
Symbolic Manipulation. Our final task is designed to measure the ability of these language
models to copy and manipulate tokens under a refined synthetic experimental setup. With the exception of GPT-J, no model exceeded 60% accuracy on
this synthetic dataset. GPT-J, by contrast, achieved
74% accuracy.
A.5 Additional Qualitative Examples
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
65
|
this synthetic dataset. GPT-J, by contrast, achieved
74% accuracy.
A.5 Additional Qualitative Examples
We provide additional qualitative examples from
our language models in Tables 21-24.
A.6 Additional Related Work
Here, we describe additional related work on different subtasks of textual style transfer that could
not be included in the main component of the paper
due to space constraints.
These works can be broadly categorized into two
families. The first family of approaches involves
identifying and replacing distinctive style-related
phrases (Li et al. (2018); Sudhakar et al. (2019);
Wu et al. (2019); Madaan et al. (2020); Malmi
et al. (2020); Reid and Zhong (2021), inter alia ).
For instance, Madaan et al. (2020) tackle the task
of politeness transfer with a two-step text-editing
approach, first identifying words with stylistic attributes using a n-gram TF-IDF method and then
training a model to replace or augment these stylistic words with ones associated with the target attribute. Similarly, Li et al. (2018) propose a simple
approach to sentiment and style transfer based on
the idea that these attributes can often be identified
by certain distinctive phrases. They identify these
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
66
|
approach to sentiment and style transfer based on
the idea that these attributes can often be identified
by certain distinctive phrases. They identify these
phrases, replace them with phrases associated with
the target attribute, and combine them with an RNN
to improve the fluency of the output text. Recently,
Reid and Zhong (2021) propose to minimize the
Levenshtein edit-distance between source and target texts, using a fine-tuned LM to make targeted
edits. In general, these approaches perform well
for very simple types of style transfer (e.g., negation by adding the word notto a sentence), but they
struggle in scenarios that require more complex
syntactic and semantic changes.
The second family of approaches involves disentangling latent representations of style and content
Hu et al. (2017); Shen et al. (2017); Fu et al. (2018);
Luo et al. (2019a); Wang et al. (2020) seek to learn
a style-invariant representation for a piece of text,
such that it can then be decoded in an arbitrary style.
For example, Hu et al. (2017) encoded sentences
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
67
|
a style-invariant representation for a piece of text,
such that it can then be decoded in an arbitrary style.
For example, Hu et al. (2017) encoded sentences
into a style-agnostic space and then decode themin a style-specific manner using a variational autoencoder alongside attribute discriminators. Shen
et al. (2017); Fu et al. (2018); Dai et al. (2019);
Wang et al. (2019) improved upon this methodology through the use of cross-alignment, style
embeddings, rule-based systems, and new architectures. While these approaches are often theoretically well-grounded, they generally require large
quantities of labeled data and struggle with scaling
beyond a small number of styles.
A.7 Computational Details
The computational cost of our experiments were
quite low, as they only involve running inference
on pre-trained models. All experiments were conducted on a single GPU. We usde an NVidia V100
for all experiments except those with GPT-J-6B,
for which we used an RTX 8000 due to memory
requirements. We estimate that all experiments for
this paper consumed fewer than 30 GPU-days.
A.8 License Details
We will release all code for this experiment under
an open-source license (MIT License).
A.9 Language Details
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
68
|
A.8 License Details
We will release all code for this experiment under
an open-source license (MIT License).
A.9 Language Details
All datasets used for this paper are in English.
A.10 Ethical Considerations
Our work aims to advance the state of research
on the task of arbitrary textual style transfer. As
with many NLP applications, these methods may
be used for negative purposes by malicious actors.
For example, it would be possible to conceive of
an instantiation of arbitrary textual style transfer
which converts a non-sensationalist news headline
into a sensationalist news headline, or one that converts a non-offensive piece of text into an offensive
piece of text, in order to achieve a malicious goal.
Our work also involves pretrained generalpurpose language models, which bring up lessobvious ethical considerations than those discussed
above. Since these language models are trained on
text scraped from the web, they have acquired some
of the biases present in web text. Such biases may
be extracted by certain forms of prompting; recent
work (Prabhumoye et al., 2021) suggests that fewshot prompts can be used to detect social biases
in pretrained language models. A large body of
work is dedicated to understanding and de-biasing
these models, but it is not the subject of our present
work.
14
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
69
|
in pretrained language models. A large body of
work is dedicated to understanding and de-biasing
these models, but it is not the subject of our present
work.
14
Dataset [Few-Shot Examples] and [Test-Time Input]
AMAZONHere is a text, which is positive: {very small but it works great in the car.} Here is a rewrite of the text, which is negative:
{very small and it works terribly in the car.} \n ### \n Here is a text, which is positive: {i really loved it and will use it alot.}
Here is a rewrite of the text, which is negative: {i really disliked it and will not use it again.} \n ### \n Here is a text, which
is positive: {it gets the job done and for the price you can t beat it.} Here is a rewrite of the text, which is negative: {it does
not work well and it was expensive.} \n ### \n Here is a text, which is negative: {i will never buy anything from this brand
again.} Here is a rewrite of the text, which is positive: {i will buy from this brand again.} \n ### \n Here is a text, which is
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
70
|
again.} Here is a rewrite of the text, which is positive: {i will buy from this brand again.} \n ### \n Here is a text, which is
negative: {if your bike had a kickstand on the plate it won’t lock down. } Here is a rewrite of the text, which is positive: {
YELPHere is a text, which is negative: {this place is awful!} Here is a rewrite of the text, which is positive: {this place is
amazing!} \n ### \n Here is a text, which is positive: {definitely will buy another pair of socks from this store–they have
the best socks ever} Here is a rewrite of the text, which is negative: {definitely will NOT buy another pair of socks from
this store–they have the worst socks ever} \n ### \n Here is a text, which is negative: {my wife and i were disappointed by
the quality of the service–also, the food was pretty tasteless} Here is a rewrite of the text, which is positive: {my wife
and i were impressive by the quality of the service–also, the food was pretty delicious} \n ### \n Here is a text, which is
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
71
|
and i were impressive by the quality of the service–also, the food was pretty delicious} \n ### \n Here is a text, which is
positive: {i loved their black tea and hot chocolate selections!} Here is a rewrite of the text, which is negative: {i hated
their black tea and hot chocolate selections!} \n ### \n Here is a text, which is positive: {it’s small yet they make you feel
right at home.} Here is a rewrite of the text, which is negative: {
SHAKESPEAREHere is a text, which is written in old English: {what hast thou there?} Here is a rewrite of the text, which is written in
modern English: {what have you got there?} \n ### \n Here is a text, which is written in old English: {what say’st thou,
my dear nurse?} Here is a rewrite of the text, which is written in modern English: {what did you say, my dear nurse?} \n
### \n Here is a text, which is written in old English: {and how doth she?} Here is a rewrite of the text, which is written in
modern English: {and how is she doing?} \n ### \n Here is a text, which is written in old English: {talk not to me, for i’ll
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
72
|
modern English: {and how is she doing?} \n ### \n Here is a text, which is written in old English: {talk not to me, for i’ll
not speak a word.} Here is a rewrite of the text, which is written in modern English: {don’t talk to me, because i won’t
answer you.} \n ### \n Here is a text, which is old English: {as mine on hers, so hers is set on mine, and all combined,
save what thou must combine by holy marriage.} Here is a rewrite of the text, which is modern English: {
GYAFC\n Here is a text, which is informal: {sorry but donnt know if i can do this alone.} Here is a rewrite of the text, which is
formal: {I am sorry, but I don’t know if I can do this alone.} \n ### \n Here is a text, which is formal: {i am going to ask
him to come to the concert with me, and i hope he accepts my invitation.} Here is a rewrite of the text, which is informal:
{gonna ask him to come to the concert with me and hope he says yes :)} \n ### \n Here is a text, which is informal: {that
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
73
|
{gonna ask him to come to the concert with me and hope he says yes :)} \n ### \n Here is a text, which is informal: {that
sucks man but u gotta move on} Here is a rewrite of the text, which is formal: {that is unfortunate, but you need to move
on} \n ### \n Here is a text, which is formal: {and i am sorry that you and your girlfriend broke up last week.} Here is a
rewrite of the text, which is informal: {and im sorry that u and ur girlfriend broke up last week...} \n ### \n Here is a text,
which is formal: {i mean that you have to really be her friend.} Here is a rewrite of the text, which is informal: {
JFLEG\n Here is a text, which is ungrammatical: {There are several reason.} Here is a rewrite of the text, which is grammatical:
{There are several reasons.} \n ### \n Here is a text, which is ungrammatical: {To my surprize nothing happened.} Here
is a rewrite of the text, which is grammatical: {To my surprise, nothing happened.} \n ### \n Here is a text, which is
ungrammatical: {This is important thing.} Here is a rewrite of the text, which is grammatical: {This is an important
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
74
|
ungrammatical: {This is important thing.} Here is a rewrite of the text, which is grammatical: {This is an important
thing.} \n ### \n Here is a text, which is ungrammatical: {Water is needed for alive.} Here is a rewrite of the text, which is
grammatical: {Water is necessary to live.} \n ### \n Here is a text, which is ungrammatical: {New and new technology
has been introduced to the society.} Here is a rewrite of the text, which is grammatical: {
SYMHere is a text, which is symbolic: {apple > seven} Here is a rewrite of the text, which is English: {apple is greater than
seven} \n ### \n Here is a text, which is symbolic: {tiger < robin} Here is a rewrite of the text, which is English: {tiger is
less than robin} \n ### \n Here is a text, which is symbolic: {teal > green} Here is a rewrite of the text, which is English:
{teal is greater than green} \n ### \n Here is a text, which is symbolic: {apple < dog} Here is a rewrite of the text, which is
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
75
|
{teal is greater than green} \n ### \n Here is a text, which is symbolic: {apple < dog} Here is a rewrite of the text, which is
English: {apple is less than dog} \n ### \n Here is a text, which is symbolic: {yellow > gray} Here is a rewrite of the text,
which is English: {
Table 8: A complete list of example-prompts used in our few-shot experiments. Here, the color gray is used to highlight the
examples used in our setups and the color teal an example test-time input in each specific TST task.
15
Positive!Negative Negative!Positive
Model Delimiter-Pair Acc r-sBLEU s-sBLEU PPL Acc r-sBLEU s-sBLEU PPL
GPT-2-Small
(117M)hi 0.35 12.4 22.7 34 0.19 11.4 23.6 33
* "" 0.43 9.0 15.9 42 0.24 7.3 14.5 40
i"" 0.46 6.6 11.3 29 0.23 6.8 14.1 30
fg 0.33 14.1 26.4 35 0.18 15.0 31.3 39
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
76
|
fg 0.33 14.1 26.4 35 0.18 15.0 31.3 39
–– 0.40 6.8 12.6 29 0.17 6.5 13.8 26
ffgg 0.36 27.0 49.7 85 0.20 27.0 56.0 94
() 0.35 18.1 32.7 54 0.18 17.6 38.2 59
"" 0.45 8.2 14.2 32 0.21 8.4 16.2 33
[] 0.35 18.9 35.5 60 0.21 14.3 29.3 43
hhhiii 0.42 6.4 12.1 24 0.19 6.7 14.1 26
GPT-2-Medium
(345M)hi 0.42 21.9 37.9 67 0.27 23.2 45.0 72
* "" 0.46 11.1 20.0 45 0.31 7.8 15.2 32
i"" 0.45 13.4 22.4 43 0.29 6.2 13.4 27
fg 0.44 21.6 38.2 73 0.26 19.1 37.1 67
–– 0.63 4.2 7.0 22 0.31 3.7 7.6 21
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
77
|
–– 0.63 4.2 7.0 22 0.31 3.7 7.6 21
ffgg 0.45 25.3 43.2 69 0.27 20.2 39.8 67
() 0.49 19.4 32.4 69 0.31 18.1 35.5 69
"" 0.47 11.3 19.2 35 0.28 9.2 17.5 34
[] 0.54 17.1 28.6 63 0.32 13.1 26.3 52
hhhiii 0.47 14.2 25.3 43 0.28 10.8 21.1 35
GPT-2-Large
(774M)hi 0.38 25.5 43.7 52 0.24 27.8 58.0 73
* "" 0.39 27.1 46.7 72 0.25 22.9 47.0 60
i"" 0.39 27.1 46.9 66 0.23 26.5 53.7 70
fg 0.39 28.7 48.8 77 0.24 28.0 54.8 63
–– 0.50 8.8 15.4 22 0.22 6.5 13.3 18
ffgg 0.43 27.7 46.5 63 0.25 36.6 69.3 113
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
78
|
ffgg 0.43 27.7 46.5 63 0.25 36.6 69.3 113
() 0.41 22.2 38.9 48 0.26 22.9 45.1 59
"" 0.52 19.7 31.4 57 0.30 17.7 34.2 48
[] 0.44 22.0 36.8 53 0.26 19.4 38.7 44
hhhiii 0.47 13.1 21.9 28 0.28 12.2 24.6 25
GPT-2-XL
(1558M)hi 0.40 26.3 43.0 81 0.31 25.9 48.2 82
* "" 0.42 22.7 39.3 60 0.29 17.6 33.3 44
i"" 0.43 20.8 35.1 54 0.29 17.9 34.0 47
fg 0.47 23.8 37.6 73 0.31 22.7 42.5 80
–– 0.56 5.6 9.0 19 0.28 4.3 7.8 18
ffgg 0.42 29.8 49.8 99 0.29 25.7 46.4 80
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
79
|
ffgg 0.42 29.8 49.8 99 0.29 25.7 46.4 80
() 0.45 16.4 28.8 41 0.28 17.8 33.6 53
"" 0.47 16.1 26.2 38 0.30 14.6 28.4 41
[] 0.46 19.2 30.8 60 0.32 16.4 31.4 52
hhhiii 0.51 9.0 13.9 25 0.37 7.4 12.7 26
GPT-Neo-1.3B
(1.3B)hi 0.48 14.9 26.1 48 0.29 11.0 21.5 40
* "" 0.41 15.1 26.8 36 0.25 13.8 28.5 44
i"" 0.38 19.7 36.0 60 0.26 18.9 37.4 48
fg 0.48 11.0 18.7 32 0.30 8.8 16.6 31
–– 0.54 5.0 8.6 18 0.30 4.9 9.6 18
ffgg 0.49 15.3 25.0 47 0.31 13.6 24.6 42
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
80
|
ffgg 0.49 15.3 25.0 47 0.31 13.6 24.6 42
() 0.44 14.3 25.3 44 0.27 15.5 29.1 51
"" 0.39 19.5 33.4 50 0.25 15.8 31.4 37
[] 0.46 15.1 27.0 51.7 0.26 15.1 30.4 47
hhhiii 0.56 5.9 9.2 28 0.32 4.6 8.1 22
GPT-Neo-2.7B
(2.7B)hi 0.43 20.0 36.1 53 0.27 22.2 43.8 59
* "" 0.37 26.2 46.7 65 0.21 26.8 54.1 65
i"" 0.37 26.1 45.9 68 0.22 23.6 50.7 60
fg 0.44 21.6 37.7 61 0.29 27.1 51.4 80
–– 0.56 4.4 7.7 15 0.23 4.0 8.2 14
ffgg 0.42 23.7 42.0 56 0.24 29.5 58.5 72
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
81
|
ffgg 0.42 23.7 42.0 56 0.24 29.5 58.5 72
() 0.44 19.7 32.9 48 0.27 21.1 40.9 64
"" 0.38 26.1 44.5 67 0.22 26.3 52.7 67
[] 0.48 20.3 35.6 67 0.25 22.4 42.9 58
hhhiii 0.45 14.1 24.8 32 0.22 21.0 42.1 55
GPT-J-6B
(6B)hi 0.40 27.3 47.0 74 0.32 17.0 32.7 51
* "" 0.38 29.2 49.5 82 0.28 23.4 42.9 61
i"" 0.36 27.4 47.2 69 0.30 23.1 43.6 64
fg 0.41 27.8 47.6 80 0.32 24.9 45.6 78
–– 0.43 7.1 12.3 19 0.20 4.8 9.0 17
ffgg 0.29 30.4 54.9 72 0.29 26.8 51.1 76
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
82
|
ffgg 0.29 30.4 54.9 72 0.29 26.8 51.1 76
() 0.48 24.9 41.6 80 0.35 22.3 39.3 77
"" 0.39 28.6 47.3 69 0.31 23.0 42.4 64
[] 0.43 23.3 38.2 63 0.37 20.8 38.3 60
hhhiii 0.34 30.3 55.6 98 0.31 23.6 44.2 67
Table 9: Zero-shot performances of the off-the-shelf “small” language models from the GPT-2 and GPT-Neo/J families on the
original version of the AMAZON dataset. Here, we also experimented with ten different delimiter-pairs, ranging from curly
brackets to asterisk quotes: Overall, curly brackets fg, square brackets [], parentheses (), and quotes ""yielded consistently
reliable and high-quality outputs. Most of the models could not go beyond 60% accuracy in the positive to negative direction and
35% accuracy in the negative to positive direction. As shown in Table 10, most models performed marginally better (in terms of
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
83
|
35% accuracy in the negative to positive direction. As shown in Table 10, most models performed marginally better (in terms of
their accuracy, BLEU, and PPL scores) on the cleaner version of the dataset, suggesting that the original version might contain
some tokenization-related (semantic) noises that might be preventing the models from performing well.
16
Positive!Negative Negative!Positive
Model Delimiter-Pair Acc r-sBLEU s-sBLEU PPL Acc r-sBLEU s-sBLEU PPL
GPT-2-Small
(117M)hi 0.34 17.9 33.4 47 0.19 13.7 30.5 42
* "" 0.43 8.1 14.8 38 0.26 7.9 16.3 37
i"" 0.45 6.3 11.6 29 0.25 7.7 15.4 32
fg 0.31 18.9 34.9 49 0.18 17.7 38.1 48
–– 0.42 6.6 12.0 24 0.21 6.6 13.7 26
ffgg 0.28 30.4 56.7 90 0.19 28.4 60.9 81
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
84
|
ffgg 0.28 30.4 56.7 90 0.19 28.4 60.9 81
() 0.34 22.2 39.1 57 0.21 18.7 40.1 48
"" 0.46 8.7 15.9 35 0.27 7.3 14.9 34
[] 0.32 19.9 35.4 59 0.20 18.4 39.7 55
hhhiii 0.39 9.3 17.0 31 0.20 9.0 18.8 28
GPT-2-Medium
(345M)hi 0.43 19.3 33.7 49 0.28 16.8 33.4 47
* "" 0.52 10.4 17.2 31 0.33 7.6 15.1 29
i"" 0.46 10.4 17.6 30 0.32 6.3 12.9 25
fg 0.48 21.9 36.7 68 0.32 20.1 38.0 57
–– 0.57 3.9 6.8 20 0.29 3.1 5.9 18
ffgg 0.45 23.3 40.1 62 0.31 21.2 41.1 64
() 0.45 19.9 33.1 50 0.32 17.6 35.3 58
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
85
|
() 0.45 19.9 33.1 50 0.32 17.6 35.3 58
"" 0.49 9.8 16.1 31 0.35 7.0 12.8 25
[] 0.50 15.7 25.9 42 0.30 13.2 26.1 46
hhhiii 0.48 8.8 14.8 26 0.31 10.1 19.2 30
GPT-2-Large
(774M)hi 0.40 26.3 44.6 68 0.29 22.9 43.4 52
* "" 0.44 23.5 40.9 54 0.27 18.8 36.9 42
i"" 0.43 20.8 36.0 44 0.29 19.4 39.2 42
fg 0.44 28.6 49.1 70 0.28 26.0 51.2 55
–– 0.40 7.3 12.0 17 0.36 7.4 13.8 19
ffgg 0.43 31.0 53.1 90 0.29 31.3 60.0 79
() 0.39 23.0 39.1 51 0.26 23.3 45.3 61
"" 0.47 18.1 29.8 47 0.31 16.8 32.7 48
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
86
|
"" 0.47 18.1 29.8 47 0.31 16.8 32.7 48
[] 0.47 20.2 34.3 50 0.26 17.5 34.1 42
hhhiii 0.49 9.9 16.2 21 0.27 9.4 18.2 22
GPT-2-XL
(1558M)hi 0.40 25.6 42.0 68 0.29 23.1 43.0 65
* "" 0.36 22.5 39.4 48 0.31 18.7 37.5 47
i"" 0.40 18.8 31.5 43 0.27 19.2 37.8 46
fg 0.46 21.5 35.4 59 0.32 22.3 41.4 70
–– 0.53 7.2 11.7 23 0.32 6.9 11.8 21
ffgg 0.45 25.7 43.2 81 0.31 24.8 45.4 72
() 0.48 19.3 30.7 52 0.30 17.8 33.4 53
"" 0.45 20.5 33.6 49 0.31 17.9 33.4 51
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
87
|
"" 0.45 20.5 33.6 49 0.31 17.9 33.4 51
[] 0.47 21.1 33.4 55 0.32 19.2 34.7 55
hhhiii 0.47 7.8 13.1 24 0.38 6.8 12.5 25
GPT-Neo-1.3B
(1.3B)hi 0.50 11.4 20.1 38 0.30 10.6 19.7 34
* "" 0.38 15.0 25.0 37 0.26 11.9 22.4 31
i"" 0.40 12.6 22.1 35 0.26 11.3 22.2 29
fg 0.49 11.8 19.9 34 0.31 10.9 20.5 35
–– 0.50 4.1 6.8 18 0.25 4.4 8.5 18
ffgg 0.48 13.9 23.9 41 0.35 12.4 22.9 38
() 0.42 16.6 27.8 53 0.28 13.1 25.8 42
"" 0.45 13.8 24.8 36 0.30 12.4 24.7 32
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
88
|
"" 0.45 13.8 24.8 36 0.30 12.4 24.7 32
[] 0.46 16.7 28.1 45 0.26 14.7 28.5 43
hhhiii 0.57 3.4 5.8 20 0.36 3.1 5.5 18
GPT-Neo-2.7B
(2.7B)hi 0.44 20.1 34.8 51 0.29 19.5 38.3 46
* "" 0.40 27.2 47.9 61 0.22 28.9 57.6 58
i"" 0.37 21.8 39.4 45 0.21 22.5 45.6 41
fg 0.48 21.4 36.5 55 0.28 23.7 45.9 57
–– 0.56 3.9 6.7 14 0.26 3.8 7.4 13
ffgg 0.43 21.2 36.2 44 0.27 28.0 55.7 56
() 0.48 17.0 28.7 42 0.32 19.5 36.8 52
"" 0.38 25.6 44.5 58 0.22 28.6 58.2 59
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
89
|
"" 0.38 25.6 44.5 58 0.22 28.6 58.2 59
[] 0.48 18.7 32.1 47 0.26 23.3 46.2 50
hhhiii 0.49 14.8 24.9 33 0.32 18.7 37.3 37
GPT-J-6B
(6B)hi 0.40 25.0 43.4 66 0.35 20.1 35.7 56
* "" 0.42 26.8 44.8 60 0.33 23.5 41.9 56
i"" 0.39 29.3 50.3 65 0.31 24.2 44.8 64
fg 0.41 26.1 44.6 57 0.33 27.1 47.7 72
–– 0.46 5.9 9.7 17 0.23 4.6 8.6 17
ffgg 0.30 29.9 53.8 56 0.30 27.5 52.3 73
() 0.44 21.0 34.8 53 0.37 19.5 37.7 64
"" 0.41 27.8 45.7 63 0.34 25.3 45.2 62
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
90
|
"" 0.41 27.8 45.7 63 0.34 25.3 45.2 62
[] 0.45 20.1 34.0 53 0.36 20.8 37.7 65
hhhiii 0.39 28.3 47.4 66 0.32 25.6 47.2 66
Table 10: Zero-shot performances of the off-the-shelf “small” language models on the clean version of the AMAZON dataset
(AMAZON -clean, in short). As before, none of the models could go beyond the 60% accuracy level, but most of them seem to
have achieved slightly better perplexity scores in the clean version of the dataset than in the original version.
17
Positive!Negative Negative!Positive
Model Delimiter-Pair Acc r-sBLEU s-sBLEU PPL Acc r-sBLEU s-sBLEU PPL
GPT-2-Small
(117M)hi 0.38 10.1 29.7 49 0.11 13.8 40.6 47
* "" 0.41 3.1 10.6 31 0.21 4.4 14.0 33
i"" 0.33 4.8 13.4 31 0.15 6.5 18.6 31
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
91
|
i"" 0.33 4.8 13.4 31 0.15 6.5 18.6 31
fg 0.36 8.4 23.3 35 0.15 7.8 23.5 34
–– 0.45 4.5 12.0 28 0.14 6.7 19.3 33
ffgg 0.37 13.4 38.9 62 0.11 16.6 49.1 54
() 0.36 8.0 25.2 42 0.13 10.9 32.7 46
"" 0.37 4.9 13.7 30 0.18 6.7 20.3 35
[] 0.36 6.8 20.7 38 0.12 8.7 25.2 31
hhhiii 0.43 2.1 5.8 17 0.09 2.1 6.4 17
GPT-2-Medium
(345M)hi 0.55 9.9 27.0 44 0.31 10.3 28.2 38
* "" 0.64 7.4 17.5 40 0.38 7.2 16.7 33
i"" 0.52 5.5 15.3 26 0.31 5.8 14.8 26
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
92
|
i"" 0.52 5.5 15.3 26 0.31 5.8 14.8 26
fg 0.66 7.6 17.6 34 0.35 8.7 24.1 34
–– 0.69 4.6 11.3 30 0.36 5.0 12.3 27
ffgg 0.68 12.2 30.3 52 0.32 14.8 36.6 58
() 0.63 8.5 22.1 46 0.32 10.5 26.8 48
"" 0.66 5.1 13.1 29 0.41 6.3 15.2 30
[] 0.66 8.4 22.0 36 0.32 8.1 21.0 33
hhhiii 0.64 1.8 4.7 16 0.24 2.2 5.9 16
GPT-2-Large
(774M)hi 0.65 14.5 36.8 54 0.22 17.3 46.7 38
* "" 0.61 13.8 33.7 43 0.27 15.8 44.4 45
i"" 0.57 16.2 44.0 59 0.27 18.7 51.4 53
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
93
|
i"" 0.57 16.2 44.0 59 0.27 18.7 51.4 53
fg 0.68 12.8 30.6 41 0.26 13.5 37.5 38
–– 0.64 8.9 22.3 31 0.24 9.7 26.3 25
ffgg 0.69 18.2 45.9 75 0.24 20.6 58.2 55
() 0.68 10.6 26.0 40 0.28 15.4 40.6 46
"" 0.74 12.0 25.9 44 0.34 14.3 34.7 42
[] 0.70 8.3 20.2 31 0.28 9.7 26.2 32
hhhiii 0.73 6.1 14.8 21 0.27 6.9 17.7 19
GPT-2-XL
(1558M)hi 0.67 15.0 35.3 59 0.41 16.4 40.4 54
* "" 0.67 10.0 25.0 35 0.37 13.0 31.6 36
i"" 0.66 10.4 25.8 41 0.34 12.5 30.2 39
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
94
|
i"" 0.66 10.4 25.8 41 0.34 12.5 30.2 39
fg 0.78 9.7 21.1 41 0.41 12.1 30.1 40
–– 0.74 6.4 13.9 25 0.37 6.3 14.2 20
ffgg 0.67 17.2 38.9 61 0.35 18.8 49.2 66
() 0.72 8.6 18.5 35 0.40 12.4 28.3 42
"" 0.72 9.7 23.3 41 0.38 10.3 24.9 34
[] 0.72 9.2 22.0 35 0.41 10.1 23.5 31
hhhiii 0.70 4.0 9.5 18 0.39 4.6 11.0 17
GPT-Neo-1.3B
(1.3B)hi 0.61 6.5 16.0 28 0.38 6.8 15.7 26
* "" 0.31 13.3 38.7 33 0.24 13.5 35.4 37
i"" 0.24 16.9 52.1 54 0.21 15.7 45.8 44
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
95
|
i"" 0.24 16.9 52.1 54 0.21 15.7 45.8 44
fg 0.66 3.2 8.4 19 0.38 5.3 12.2 21
–– 0.52 2.9 8.4 17 0.30 4.4 11.2 20
ffgg 0.60 9.1 23.6 35 0.39 8.5 21.2 30
() 0.59 6.8 18.6 34 0.27 11.1 31.1 47
"" 0.46 14.9 40.2 54 0.23 14.9 40.1 47
[] 0.57 8.1 20.8 38 0.36 8.4 22.1 33
hhhiii 0.68 1.3 3.8 17 0.38 1.9 4.1 16
GPT-Neo-2.7B
(2.7B)hi 0.68 8.3 21.8 28 0.28 12.5 33.1 31
* "" 0.56 14.1 39.1 54 0.18 18.1 51.9 51
i"" 0.54 12.1 34.3 43 0.21 15.9 46.4 42
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
2205.11503
|
96
|
i"" 0.54 12.1 34.3 43 0.21 15.9 46.4 42
fg 0.63 7.5 19.5 27 0.26 12.4 32.8 32
–– 0.63 3.4 8.5 16 0.26 3.7 9.8 14
ffgg 0.55 12.0 32.9 40 0.20 16.2 48.1 42
() 0.73 7.3 18.0 38 0.34 13.6 34.2 47
"" 0.55 12.6 33.2 42 0.17 15.4 45.9 42
[] 0.62 8.0 21.3 33 0.27 13.7 37.0 39
hhhiii 0.64 4.6 12.6 21 0.27 7.1 18.5 20
GPT-J-6B
(6B)hi 0.60 14.9 37.6 52 0.44 14.4 33.0 51
* "" 0.57 16.2 41.0 57 0.36 16.1 37.1 49
i"" 0.52 16.4 43.3 62 0.32 17.2 42.6 56
|
2205.11503
|
Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models
|
We propose a method for arbitrary textual style transfer (TST)--the task of
transforming a text into any given style--utilizing general-purpose pre-trained
language models. Our method, Prompt-and-Rerank, is based on a mathematical
formulation of the TST task, decomposing it into three constituent components:
textual similarity, target style strength, and fluency. Specifically, our
method first uses zero-shot or few-shot prompting to obtain a set of candidate
generations in the target style, and then re-ranks these candidates according
to a combination of the three components above. Empirically, our method enables
small pre-trained language models to perform on par with state-of-the-art
large-scale models while consuming two orders of magnitude less compute and
memory. Finally, we conduct a systematic investigation of the effect of model
size and prompt design (e.g., prompt paraphrasing and delimiter-pair choice) on
style transfer quality across seven diverse textual style transfer datasets.
|
http://arxiv.org/pdf/2205.11503
|
[
"Mirac Suzgun",
"Luke Melas-Kyriazi",
"Dan Jurafsky"
] |
[
"cs.CL"
] |
GitHub page: https://github.com/suzgunmirac/prompt-and-rerank.
Project page: https://lukemelas.github.io/prompt-and-rerank/
| null |
cs.CL
|
20220523
|
20220523
|
[
{
"id": "2110.10668"
},
{
"id": "2105.08206"
},
{
"id": "1904.02295"
},
{
"id": "2112.07868"
},
{
"id": "1905.10060"
},
{
"id": "2205.11503"
},
{
"id": "2010.01054"
},
{
"id": "1908.08039"
}
] |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.