SEBSem: simple and efficient biomedical semantic relatedness measure

Abstract

Calculating semantic relatedness between terms is crucial in numerous knowledge and information processing tasks highly relevant to the biomedical domain. Examples include semantic search and automated processing of scientific texts. Most available methods rely heavily on highly specialised resources, which substantially limits their reusability in various applications within the domain. In this work we present a simple semantic relatedness measure that relies only on very general resources and its design features allow minimising the costs of online computations. The relatedness is computed through comparing automatically extracted key-phrases relevant to respective input terms. This simple strategy provides a method that gives promising early test results, comparable to those of human annotators and state-of-the-art methods, on a well established benchmark.