MedDialog: Large-scale Medical Dialogue Datasets

Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, pages 9241–9250, November 16–20, 2020. c⃝2020 Association for Computational Linguistics 9241 MedDialog: Large-scale Medical Dialogue Datasets Guangtao Zeng, Wenmian Yang, Zeqian Ju, Yue Yang, Sicheng Wang, Ruisi Zhang, Meng Zhou, Jiaqi Zeng, Xiangyu Dong, Ruoyu Zhang, Hongchao Fang, Penghui Zhu, Shu Chen and Pengtao Xie UC San Diego p1xie@eng.ucsd.edu Abstract Medical dialogue systems are promising in as- sisting in telemedicine to increase access to healthcare services, improve the quality of pa- tient care, and reduce medical costs. To facil- itate the research and development of medical dialogue systems, we build large-scale med- ical dialogue datasets – MedDialog, which contain 1) a Chinese dataset with 3.4 mil- lion conversations between patients and doc- tors, 11.3 million utterances, 660.2 million tokens, covering 172 specialties of diseases, and 2) an English dataset with 0.26 million conversations, 0.51 million utterances, 44.53 million tokens, covering 96 specialties of dis- eases. To our best knowledge, MedDialog is the largest medical dialogue dataset to date. We pretrain several dialogue generation mod- els on the Chinese MedDialog dataset, includ- ing Transformer, GPT, BERT-GPT, and com- pare their performance. It is shown that mod- els trained on MedDialog are able to gener- ate clinically correct and human-like medi- cal dialogues. We also study the transferabil- ity of models trained on MedDialog to low- resource medical dialogue generation tasks. It is shown that via transfer learning which fine- tunes the models pretrained on MedDialog, the performance on medical dialogue generation tasks with small datasets can be greatly im- proved, as shown in human evaluation and automatic evaluation. The datasets and code are available at https://github.com/UCSD- AI4H/Medical-Dialogue-System 1 Introduction Telemedicine refers to the practice of delivering pa- tient care remotely, where doctors provide medical consultations to patients using HIPAA compliant video-conferencing tools. As an important comple- ment to traditional face-to-face medicine practiced physically in hospitals and clinics, telemedicine has a number of advantages. First, it increases access to care. For people living in medically under-served communities (e.g., rural areas) that are in shortage of clinicians, telemedicine enables them to receive faster and cheaper care compared with traveling over a long distance to visit a clinician. Second, it reduces healthcare costs. In a study1 by Jeffer- son Health, it is shown that diverting patients from emergency departments with telemedicine can save more than $1,500 per visit. Third, telemedicine can improve the quality of care. The study in (Pande and Morris, 2015) shows that telemedicine patients score lower for depression, anxiety, and stress, and have 38% fewer hospital admissions. Other advan- tages include improving patient engagement and satisfaction, improving provider satisfaction, etc. Please refer to (Wootton et al., 2017) for a more comprehensive review. While telemedicine is promising, it has sev- eral limitations. First, it puts additional burden on physicians. In addition to practicing face-to- face medicine which already makes physicians very busy, physicians need to provide remote telemedicine consultations, which further increases the risk of physician burnout. Second, different from …

Content truncated to 500 words