Learning to Generate Explanation from e-Hospital Services for Medical Suggestion

Proceedings of the 29th International Conference on Computational Linguistics, pages 2946–2951 October 12–17, 2022. 2946 Learning to Generate Explanation from e-Hospital Services for Medical Suggestion Wei-Lin Chen,1 An-Zi Yen,2 Hen-Hsen Huang,3 Hsin-Hsi Chen1 1Department of Computer Science and Information Engineering, National Taiwan University, Taiwan 2Department of Computer Science, National Yang Ming Chiao Tung University, Taiwan 3Institute of Information Science, Academia Sinica, Taiwan wlchen@nlg.csie.ntu.edu.tw, azyen@nycu.edu.tw, hhhuang@iis.sinica.edu.tw, hhchen@ntu.edu.tw Abstract Explaining the reasoning of neural models has attracted attention in recent years. Providing highly-accessible and comprehensible expla- nations in natural language is useful for hu- mans to understand the model’s prediction re- sults. In this work, we present a pilot study to investigate explanation generation with a narrative and causal structure for the scenario of health consulting. Our model generates a medical suggestion regarding the patient’s con- cern and provides an explanation as the outline of the reasoning. To align the generated ex- planation with the suggestion, we propose a novel discourse-aware mechanism with multi- task learning. Experimental results show that our model achieves promising performances in both quantitative and human evaluation. 1 Introduction Neural models have shown remarkable success in various tasks, however, simply offering the predic- tions may not satisfy the requirement of end-users. Understanding how the decision has been reached by the model is essential in real-world applications. To provide a meaningful, human-comprehensible explanation, presenting it in natural language is a proper fashion. Note that simply present the ex- planation as a shopping list or fragments of text- highlight is not an ideal way. Humans prefer to read a text composed of a narrative structure, or- ganized by the discourse relations elaborating the causality between the model input and output (Re- iter, 2019). In this work, we propose a novel health consultancy model which can provide medical sug- gestions accompanied with natural language expla- nations learned from medical specialists to help humans make decisions in their daily life. As many people are eager for help in addressing their health concerns, a model is necessitated to be capable of not only providing suggestions re- garding their concerns, but also explaining the sug- gestions, alleviating their worries before visiting male / (20~29), 2021/02/25 / Ask： I went to the ER in April after having a fever, the initial diagnosis is urethritis. The doctor gave me antibiotics and antipyretics, however, my temperature didn’t stepped down and I went to another hospital for examinations (a blood test and urine test). The results came back normal but the temperature kept running high. How come I am still having a fever? Are there any further examinations I should take? OO hospital OOO M.D. / Internal Medicine Department / 2021/02/25 / Response： It seems that you have been feeling feverishness for over a month. To evaluate the possibility that the fever is not caused by infections and assess the need for further referrals, I suggest that you make an appointment with doctors of infectious disease department. Figure 1: A Data Instance from the Health Consultancy Website the …

Content truncated to 500 words