Dr. Summarize: Global Summarization of Medical Dialogue by Exploiting Local Structures.

Findings of the Association for Computational Linguistics: EMNLP 2020, pages 3755–3763 November 16 - 20, 2020. c⃝2020 Association for Computational Linguistics 3755 Dr. Summarize: Global Summarization of Medical Dialogue by Exploiting Local Structures Anirudh Joshi Stanford University anirudhjoshi@stanford.edu Namit Katariya Curai Xavier Amatriain Curai Anitha Kannan Curai Abstract Understanding a medical conversation be- tween a patient and a physician poses unique natural language understanding chal- lenge since it combines elements of standard open-ended conversation with very domain- specific elements that require expertise and medical knowledge. Summarization of med- ical conversations is a particularly important aspect of medical conversation understanding since it addresses a very real need in medical practice: capturing the most important aspects of a medical encounter so that they can be used for medical decision making and subsequent follow ups. In this paper we present a novel approach to medical conversation summarization that leverages the unique and independent local structures created when gathering a patient’s medical history. Our approach is a variation of the pointer generator network where we in- troduce a penalty on the generator distribu- tion, and we explicitly model negations. The model also captures important properties of medical conversations such as medical knowl- edge coming from standardized medical on- tologies better than when those concepts are introduced explicitly. Through evaluation by doctors, we show that our approach is pre- ferred on twice the number of summaries to the baseline pointer generator model and cap- tures most or all of the information in 80% of the conversations making it a realistic alterna- tive to costly manual summarization by medi- cal experts. 1 Introduction Telemedicine is a rapidly growing medium of in- teraction with the healthcare system (Mann et al., 2020). With the COVID-19 pandemic limiting in person medical visits, healthcare systems are see- ing greater than 100% increase in virtual urgent care visits and greater than 4000% increase in DR: You mentioned having a cough for 2 days and a fever since last night along with being short of breath. Is that correct? PT: yes , correct DR: I appreciate your concern for preventing spread. Do you feel like you are unable to move around as usual? PT: I’m definitely weaker and low energy the fever went down to 99 this morning DR: Have you taken any medications or tried anything else to help you with your symptoms? PT: lots of fluids and vitamin c. lozenges to minimize coughing DR: do you have any medical conditions or have you been on any medications PT: no, none DR: alright. When you had a fever, did you take medicine like tylenol to bring the fever down? PT: I didn’t Model Output Summary • mentioned having a cough for 2 days and a fever since last night along with being short of breath. • unable to move around as usual. definitely weaker and low energy fever went down to 99 • lots of fluids and vitamin c. lozenges to mini- mize coughing with symptoms . • no medical conditions. …

Content truncated to 500 words