ODSQA (Open-Domain Spoken Question Answering Dataset) From ODSQA: Open-domain Spoken Question Answering Dataset This is a Chinese dataset. In addition, an English dataset Spoken-SQuAD is also provided.hereturn up. Spoken-SQuAD  It is a spoken intelligent question-answering corpus generated from the SQuAD dataset through Google's text-to-speech (TTS) system. Although Spoken-SQuAD is large enough to train the most advanced intelligent question-answering models, it is artificially generated, so there is still a certain gap with real spoken question-answering. Therefore, researchers released an SQA dataset containing more than three thousand questions, called ODSQA. It is currently the largest real SQA dataset for extraction-based intelligent question-answering tasks.

Citation

@article{lee2018spoken, title={Spoken SQuAD: A Study of Mitigating the Impact of Speech Recognition Errors on Listening Comprehension}, author={Lee, Chia-Hsuan and Wu, Szu-Lin and Liu, Chi-Liang and Lee, Hung-yi}, journal={Proc. Interspeech 2018}, pages={3459–3463}, year={2018} }