CCQA: A New Web-Scale Question Answering Dataset for Model Pre-Training

Patrick Huber, Armen Aghajanyan, Barlas Oguz, Dmytro Okhonko, Scott Yih, Sonal Gupta, Xilun Chen


Abstract
We propose a novel open-domain question-answering dataset based on the Common Crawl project. With a previously unseen number of around 130 million multilingual question-answer pairs (including about 60 million English data-points), we use our large-scale, natural, diverse and high-quality corpus to in-domain pre-train popular language models for the task of question-answering. In our experiments, we find that our Common Crawl Question Answering dataset (CCQA) achieves promising results in zero-shot, low resource and fine-tuned settings across multiple tasks, models and benchmarks.
Anthology ID:
2022.findings-naacl.184
Volume:
Findings of the Association for Computational Linguistics: NAACL 2022
Month:
July
Year:
2022
Address:
Seattle, United States
Editors:
Marine Carpuat, Marie-Catherine de Marneffe, Ivan Vladimir Meza Ruiz
Venue:
Findings
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
2402–2420
Language:
URL:
https://aclanthology.org/2022.findings-naacl.184
DOI:
10.18653/v1/2022.findings-naacl.184
Bibkey:
Cite (ACL):
Patrick Huber, Armen Aghajanyan, Barlas Oguz, Dmytro Okhonko, Scott Yih, Sonal Gupta, and Xilun Chen. 2022. CCQA: A New Web-Scale Question Answering Dataset for Model Pre-Training. In Findings of the Association for Computational Linguistics: NAACL 2022, pages 2402–2420, Seattle, United States. Association for Computational Linguistics.
Cite (Informal):
CCQA: A New Web-Scale Question Answering Dataset for Model Pre-Training (Huber et al., Findings 2022)
Copy Citation:
PDF:
https://aclanthology.org/2022.findings-naacl.184.pdf
Video:
 https://aclanthology.org/2022.findings-naacl.184.mp4
Code
 facebookresearch/CCQA
Data
CCQACC100CCNetELI5GooAQNatural QuestionsTriviaQA