Comparative Analysis of Semantic Similarity Techniques for Medical Text
Fakhare Alam, Muhammad Afzal, Khalid Mahmood Malik · 2020
In medical domain, physicians frequently consult information and knowledge sources for learning relevant piece of information for their clinical decisions. The sources are either internal such as electronic medical records databases or external sources such as biomedical literature. The key challenge for medical information extraction systems is finding required information, considering the explicit and implicit clinical context, with high precision and recall. Previously, many approaches were presented to exploit different aspects of information in the user query and medical text for obtaining semantic similarity at a concept, sentence, and document level, however, there doesn't exist a comparative analysis of different semantic similarity approaches for biomedical domain. This paper aims to compare different techniques used for finding semantic similarity by performing empirical analysis. We have implemented various modern semantic similarity techniques including ontological based matching and contextual word embedding.