Chatbot based on clinical literature for decision support

Abstract

Clinical practice guidelines try to provide the state-of-the-art in diagnostic and treatment methods for each disease, by a systematic review of the scientific evidence, but it can be difficult to keep up to date in a context of healthcare in constant evolution. Improvements in Deep Learning and Natural Language Processing have allowed to perform multiple applications, such as conversational agents (chatbots or virtual assistants), that are designed to simulate a human conversation. Language models behind these systems are able to analyze a huge collection of documents with unstructured data and extract the essential information from each one, easing the fast consultation of guidelines by practitioners and patients. This article provide an approach of a thesis plan to analyze different techniques and language models, and develop a chatbot able to answer according to clinical practice guidelines and other high-quality biomedical literature in a real-time decision support system for healthcare professionals, patients, and caregivers.