CDialog: A Multi-turn Covid-19 Conversation Dataset for Entity-Aware Dialog Generation

Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 11373 - 11385 December 7-11, 2022 ©2022 Association for Computational Linguistics CDialog: A Multi-turn Covid-19 Conversation Dataset for Entity-Aware Dialog Generation Deeksha Varshney†∗, Aizan Zafar†*, Niranshu Kumar Behra† Asif Ekbal† †Department of Computer Science and Engineering, Indian Institute of Technology Patna, India {1821cs13,aizan_1921cs17,niranshu_1901cs39,asif}@iitp.ac.in Abstract The development of conversational agents to interact with patients and deliver clinical advice has attracted the interest of many researchers, particularly in light of the COVID-19 pandemic. The training of an end-to-end neural based dia- log system, on the other hand, is hampered by a lack of multi-turn medical dialog corpus. We make the very first attempt to release a high- quality multi-turn Medical Dialog dataset relat- ing to Covid-19 disease named CDialog, with over 1K conversations collected from the on- line medical counselling websites. We annotate each utterance of the conversation with seven different categories of medical entities, includ- ing diseases, symptoms, medical tests, medical history, remedies, medications and other as- pects as additional labels. Finally, we propose a novel neural medical dialog system based on the CDialog dataset to advance future re- search on developing automated medical dia- log systems. We use pre-trained language mod- els for dialogue generation, incorporating an- notated medical entities, to generate a virtual doctor’s response that addresses the patient’s query. Experimental results show that the pro- posed dialog models perform comparably bet- ter when supplemented with entity information and hence can improve the response quality. 1 Introduction Currently, telemedicine is absolutely appropriate in reducing the risk of COVID-19 among health- care providers and patients due to the diversion of medical resources as millions of people around the world have experienced delays in diagnosis and treatment. Conversational agents (Gopalakr- ishnan et al., 2019; Zhao et al., 2020; Wu et al., 2018; Reddy et al., 2019) have been proved to be effective in carrying on a natural conversation and understanding the meanings of words to respond with a coherent dialog. It has been also effective in ∗Equal Conrtibution providing support to complete several tasks such as booking a ticket (Liao et al., 2020), getting reser- vations (Wei et al., 2018), etc. In medical domain, (Zeng et al., 2020; Liu et al., 2020; Li et al., 2021; Wei et al., 2018; Xu et al., 2019) have come up with standard techniques to model medical dialogs which reduces face-to-face consultations, resulting in reduced costs and helps the patient get quicker medical treatment. However, medical dialog sys- tems are more difficult to implement than the stan- dard task-oriented dialog systems (TDSs) as there are several other professional phrases / formal med- ical expressions that are frequently conveyed while communicating (Shi et al., 2020). A significant effort has recently been undertaken to collect medical dialog data for research on med- ical dialog systems (Shi et al., 2020; Zeng et al., 2020; Liao et al., 2020; Liu et al., 2020; Yang et al., 2020). They all, however, have some limitations: (i) A comprehensive …

Content truncated to 500 words