Rijgersberg/common_corpus_nl

Common Corpus v2 NL

This is a version of Common Corpus v2 filtered to keep only the rows where language is "Dutch".

Common Corpus is a very large open and permissible licensed text dataset created by Pleias. Please be sure to acknowledge the creators of the original dataset when using this filtered version.

Filtering

Common Corpus is a collection of disparate datasets. Note that filtering the entire collection for rows where the language is "Dutch" is not the same as filtering for entire datasets that are supposed to be Dutch-language. Since language classification is an automated process, there may be false positives and false negatives.

Examples are:

• false positives: code from StackExchange, French from French-PD-diverse are included in Common Corpus NL when they shouldn't be.
• false negatives: rows from Dutch-PD that were misclassified, for example as English, are not included in Common Corpus NL when they should be.

If you want to use entire datasets, you either have to look up what the source was for including it into Common Corpus, or filter Common Corpus yourself.

Usage

from datasets import load_dataset

# load the full dataset
dataset = load_dataset('Rijgersberg/common_corpus_nl', 'all', split='train')

# load only a specific subset
wikipedia = load_dataset('Rijgersberg/common_corpus_nl', 'Open Web-Wikipedia', split='train')

Contents

The dataset has the following content. Tokens are measured by the robbert-2023-dutch-base Dutch tokenizer on the text column only. word_count is taken directly from the word_count column of the dataset.

collection	open_type	row_count	word_count	token_count
Dutch-PD	Open Culture	198,090	1,341,547,229	2,453,085,804
French-PD-diverse	Open Culture	5,357	38,359,009	75,965,219
Gutenberg	Open Culture	4,844	44,405,784	74,446,253
US-PD-Books	Open Culture	3,714	23,298,108	53,772,090
Multilingual-PD	Open Culture	3,164	22,935,496	45,033,262
English-PD	Open Culture	2,622	18,344,861	38,407,512
German-PD	Open Culture	1,744	12,434,468	26,319,336
US-PD-Newspapers	Open Culture	1,478	5,297,259	11,335,348
Latin-PD	Open Culture	637	4,406,870	9,623,548
LoC-PD-Books	Open Culture	480	3,384,886	6,359,398
Italian-PD	Open Culture	253	1,767,994	4,185,808
French-PD-Books	Open Culture	195	1,462,312	3,326,121
Europeana	Open Culture	126	751,961	2,018,302
Spanish-PD-Books	Open Culture	114	819,389	1,831,298
French-PD-Newspapers	Open Culture	117	589,899	1,352,344
Danish-PD	Open Culture	36	260,479	578,291
Spanish-PD-Newspapers	Open Culture	34	221,839	533,701
German-PD-Newspapers	Open Culture	27	155,979	398,229
NewZealand-PD-Newspapers	Open Culture	70	135,552	348,682
Polish-PD	Open Culture	6	37,573	135,721
Portuguese-PD	Open Culture	5	35,666	84,105
Greek-PD	Open Culture	1	11,084	23,343
BNL Newspapers (1841-1879)	Open Culture	42	6,409	17,658
Wikisource	Open Culture	1	36	58
Eurlex	Open Government	269,340	948,031,137	2,104,235,522
Eurovoc	Open Government	46,006	480,170,115	971,763,805
French Open Data	Open Government	228,097	211,210,103	546,913,250
Marianne-Europe	Open Government	10,101	44,308,181	113,369,293
TEDEUTenders	Open Government	5,105	3,423,351	8,123,266
USPTO	Open Government	40	410,010	1,505,353
UN-Digital-Library	Open Government	19	95,691	436,028
WTO	Open Government	11	55,733	125,785
Court Listener	Open Government	243	22,176	82,334
OECD	Open Government	5	30,895	56,160
Caselaw Access Project	Open Government	450	13,344	33,405
GATT_library	Open Government	1	282	1,112
OpenAlex	Open Science	11,867	76,200,223	142,457,454
French-Science-Pile	Open Science	1,485	6,197,572	18,685,715
Open-Science-Pile	Open Science	1,199	4,711,962	8,011,769
German-Science-Pile	Open Science	985	4,234,708	7,488,555
Spanish-Science-Pile	Open Science	163	1,071,826	2,263,934
Wikipedia	Open Web	2,135,977	367,689,443	634,440,794
StackExchange	Open Web	270,147	117,494,333	464,336,069
Youtube-Commons	Open Web	1,982	5,886,772	8,329,426
None	None	2	729	1,685
total		3,206,382	3,791,928,728	7,841,842,145

Code

In principle it is very easy to create Common Corpus NL by filtering Common Corpus using Hugging Face datasets' dataset.filter() functionality. However, Common Corpus is larger than my available disk space.

A possible solution is to process Common Corpus streaming and relying on the fact that the Dutch subset will be much, much smaller than the full dataset. The code for that solution is below. However, I was having trouble streaming the entire dataset without any connection errors along the way.

from datasets import load_dataset, Dataset


common_corpus = load_dataset('PleIAs/common_corpus', split='train', streaming=True)

def nl():
    for row in common_corpus:
        if row['language'] == 'Dutch':
            yield row

common_corpus_nl = Dataset.from_generator(nl)

common_corpus_nl.push_to_hub('Rijgersberg/common_corpus_nl')

Therefore I took the approach for every one of ten subfolders of Common Corpus:

• downloading the subfolder in a fault-tolerant way
• doing the filtering to Dutch rows only
• uploading that by itself to the Hugging Face hub
• deleting all the downloaded files and datasets cache files (around 1.5 TB for every subfolder)

Then finally I concatenated the ten Dutch datasets into a single one, which is the one you are looking at.

import shutil

from datasets import concatenate_datasets, load_dataset
from huggingface_hub import snapshot_download
from huggingface_hub.errors import LocalEntryNotFoundError
from requests import ReadTimeout


local_dir = '/path/to/downloadfolder/commoncorpus'

for i in range(1, 10+1):
    success = False
    while not success:
        try:
            # download one common corpus folder at a time to a local directory
            snapshot_download(  # will skip files that have already been downloaded
                repo_id='PleIAs/common_corpus',
                repo_type='dataset',
                allow_patterns=f'common_corpus_{i}/*',
                local_dir=local_dir
            )
            success = True
        except (LocalEntryNotFoundError, ReadTimeout) as e:
            print(e)

    subdataset = load_dataset(local_dir, split='train')
    subdataset = subdataset.filter(lambda lang: lang == 'Dutch', input_columns=['language'])

    subdataset.push_to_hub(f'Rijgersberg/common_corpus_nl_{i}')

    # remove local copies of the data to free up disk space
    shutil.rmtree(local_dir)
    shutil.rmtree('/path/to/cache/huggingface/datasets/commoncorpus')

# concatenate all (much smaller) Dutch datasets into a single dataset
common_corpus_nl = concatenate_datasets([load_dataset(f'Rijgersberg/common_corpus_nl_{i}', split='train')
                                         for i in range(1, 10+1)])
common_corpus_nl.push_to_hub('Rijgersberg/common_corpus_nl')