alexandrainst/scandi-qa

Dataset Card for ScandiQA

Dataset Summary

ScandiQA is a dataset of questions and answers in the Danish, Norwegian, and Swedish languages. All samples come from the Natural Questions (NQ) dataset, which is a large question answering dataset from Google searches. The Scandinavian questions and answers come from the MKQA dataset, where 10,000 NQ samples were manually translated into, among others, Danish, Norwegian, and Swedish. However, this did not include a translated context, hindering the training of extractive question answering models.

We merged the NQ dataset with the MKQA dataset, and extracted contexts as either "long answers" from the NQ dataset, being the paragraph in which the answer was found, or otherwise we extract the context by locating the paragraphs which have the largest cosine similarity to the question, and which contains the desired answer.

Further, many answers in the MKQA dataset were "language normalised": for instance, all date answers were converted to the format "YYYY-MM-DD", meaning that in most cases these answers are not appearing in any paragraphs. We solve this by extending the MKQA answers with plausible "answer candidates", being slight perturbations or translations of the answer.

With the contexts extracted, we translated these to Danish, Swedish and Norwegian using the DeepL translation service for Danish and Swedish, and the Google Translation service for Norwegian. After translation we ensured that the Scandinavian answers do indeed occur in the translated contexts.

As we are filtering the MKQA samples at both the "merging stage" and the "translation stage", we are not able to fully convert the 10,000 samples to the Scandinavian languages, and instead get roughly 8,000 samples per language. These have further been split into a training, validation and test split, with the latter two containing roughly 750 samples. The splits have been created in such a way that the proportion of samples without an answer is roughly the same in each split.

Supported Tasks and Leaderboards

Training machine learning models for extractive question answering is the intended task for this dataset. No leaderboard is active at this point.

Languages

The dataset is available in Danish (da), Swedish (sv) and Norwegian (no).

Dataset Structure

Data Instances

• Size of downloaded dataset files: 69 MB
• Size of the generated dataset: 67 MB
• Total amount of disk used: 136 MB

An example from the train split of the da subset looks as follows.

{
    'example_id': 123,
    'question': 'Er dette en test?',
    'answer': 'Dette er en test',
    'answer_start': 0,
    'context': 'Dette er en testkontekst.',
    'answer_en': 'This is a test',
    'answer_start_en': 0,
    'context_en': "This is a test context.",
    'title_en': 'Train test'
}

Data Fields

The data fields are the same among all splits.

• example_id: an int64 feature.
• question: a string feature.
• answer: a string feature.
• answer_start: an int64 feature.
• context: a string feature.
• answer_en: a string feature.
• answer_start_en: an int64 feature.
• context_en: a string feature.
• title_en: a string feature.

Data Splits

name	train	validation	test
da	6311	749	750
sv	6299	750	749
no	6314	749	750

Dataset Creation

Curation Rationale

The Scandinavian languages does not have any gold standard question answering dataset. This is not quite gold standard, but the fact both the questions and answers are all manually translated, it is a solid silver standard dataset.

Source Data

The original data was collected from the MKQA and Natural Questions datasets from Apple and Google, respectively.

Additional Information

Dataset Curators

Dan Saattrup Nielsen from the The Alexandra Institute curated this dataset.

Licensing Information

The dataset is licensed under the CC BY-SA 4.0 license.