Title: Semantic Similarity To Improve Question Understanding in a Virtual Patient

Abstract: In medicine, a communicating virtual patient or doctor allows students to train in medical diagnosis and develop skills to conduct a medical consultation. In this paper, we describe a conversational virtual standardized patient system to allow medical students to simulate a diagnosis strategy of an abdominal surgical emergency. We exploited the semantic properties captured by distributed word representations to search for similar questions in the virtual patient dialogue system. We created two dialogue systems that were evaluated on datasets collected during tests with students. The first system based on hand-crafted rules obtains $92.29\%$ as $F1$-score on the studied clinical case while the second system that combines rules and semantic similarity achieves $94.88\%$. It represents an error reduction of $9.70\%$ as compared to the rules-only-based system.