Dataset Card for "search_qa"

Dataset Summary

We publicly release a new large-scale dataset, called SearchQA, for machine comprehension, or question-answering. Unlike recently released datasets, such as DeepMind CNN/DailyMail and SQuAD, the proposed SearchQA was constructed to reflect a full pipeline of general question-answering. That is, we start not from an existing article and generate a question-answer pair, but start from an existing question-answer pair, crawled from J! Archive, and augment it with text snippets retrieved by Google. Following this approach, we built SearchQA, which consists of more than 140k question-answer pairs with each pair having 49.6 snippets on average. Each question-answer-context tuple of the SearchQA comes with additional meta-data such as the snippet's URL, which we believe will be valuable resources for future research. We conduct human evaluation as well as test two baseline methods, one simple word selection and the other deep learning based, on the SearchQA. We show that there is a meaningful gap between the human and machine performances. This suggests that the proposed dataset could well serve as a benchmark for question-answering.

Supported Tasks and Leaderboards

More Information Needed

Languages

More Information Needed

Dataset Structure

Data Instances

raw_jeopardy

• Size of downloaded dataset files: 3.31 GB
• Size of the generated dataset: 7.77 GB
• Total amount of disk used: 11.09 GB

An example of 'train' looks as follows.

train_test_val

• Size of downloaded dataset files: 3.15 GB
• Size of the generated dataset: 7.51 GB
• Total amount of disk used: 10.66 GB

An example of 'validation' looks as follows.

Data Fields

The data fields are the same among all splits.

raw_jeopardy

• category: a string feature.
• air_date: a string feature.
• question: a string feature.
• value: a string feature.
• answer: a string feature.
• round: a string feature.
• show_number: a int32 feature.
• search_results: a dictionary feature containing:
  • urls: a string feature.
  • snippets: a string feature.
  • titles: a string feature.
  • related_links: a string feature.

train_test_val

• category: a string feature.
• air_date: a string feature.
• question: a string feature.
• value: a string feature.
• answer: a string feature.
• round: a string feature.
• show_number: a int32 feature.
• search_results: a dictionary feature containing:
  • urls: a string feature.
  • snippets: a string feature.
  • titles: a string feature.
  • related_links: a string feature.

Data Splits

raw_jeopardy

train_test_val

Dataset Creation

Curation Rationale

More Information Needed

Source Data

Initial Data Collection and Normalization

More Information Needed

Who are the source language producers?

More Information Needed

Annotations

Annotation process

More Information Needed

Who are the annotators?

More Information Needed

Personal and Sensitive Information

More Information Needed

Considerations for Using the Data

Social Impact of Dataset

More Information Needed

Discussion of Biases

More Information Needed

Other Known Limitations

More Information Needed

Additional Information

Dataset Curators

More Information Needed

Licensing Information

More Information Needed

Citation Information

@article{DBLP:journals/corr/DunnSHGCC17,
    author    = {Matthew Dunn and
                Levent Sagun and
                Mike Higgins and
                V. Ugur G{"{u}}ney and
                Volkan Cirik and
                Kyunghyun Cho},
    title     = {SearchQA: {A} New Q{\&}A Dataset Augmented with Context from a
                Search Engine},
    journal   = {CoRR},
    volume    = {abs/1704.05179},
    year      = {2017},
    url       = {http://arxiv.org/abs/1704.05179},
    archivePrefix = {arXiv},
    eprint    = {1704.05179},
    timestamp = {Mon, 13 Aug 2018 16:47:09 +0200},
    biburl    = {https://dblp.org/rec/journals/corr/DunnSHGCC17.bib},
    bibsource = {dblp computer science bibliography, https://dblp.org}
}

Contributions

Thanks to @lewtun, @mariamabarham, @lhoestq, @thomwolf for adding this dataset.