Multilingual Medical Question Answering and Information Retrieval for Rural Health Intelligence Access

In rural regions of several developing countries, access to quality\nhealthcare, medical infrastructure, and professional diagnosis is largely\nunavailable. Many of these regions are gradually gaining access to internet\ninfrastructure, although not with a strong enough connection to allow for\nsustained communication with a medical practitioner. Several deaths resulting\nfrom this lack of medical access, absence of patient's previous health records,\nand the unavailability of information in indigenous languages can be easily\nprevented. In this paper, we describe an approach leveraging the phenomenal\nprogress in Machine Learning and NLP (Natural Language Processing) techniques\nto design a model that is low-resource, multilingual, and a preliminary\nfirst-point-of-contact medical assistant. Our contribution includes defining\nthe NLP pipeline required for named-entity-recognition, language-agnostic\nsentence embedding, natural language translation, information retrieval,\nquestion answering, and generative pre-training for final query processing. We\nobtain promising results for this pipeline and preliminary results for EHR\n(Electronic Health Record) analysis with text summarization for medical\npractitioners to peruse for their diagnosis. Through this NLP pipeline, we aim\nto provide preliminary medical information to the user and do not claim to\nsupplant diagnosis from qualified medical practitioners. Using the input from\nsubject matter experts, we have compiled a large corpus to pre-train and\nfine-tune our BioBERT based NLP model for the specific tasks. We expect recent\nadvances in NLP architectures, several of which are efficient and\nprivacy-preserving models, to further the impact of our solution and improve on\nindividual task performance.\n