Discovering Semantic Relationships between Concepts from MEDLINE

Abstract

This paper presents an approach for detecting unapparent links between two concepts from MEDLINE. Given two topics A and C, for example, two biomedical concepts fish oil and raynauds, we attempt to find the best concept chain and evidence trail that connect them across documents. The technique begins with developing concept profiles for each input topic, and then grouping relevant concepts by semantic types, and finally generating ranked concept chains and evidence trails connecting topics. Biomedical corpus from MEDLINE is used to evaluate the performance of the system and demonstrates the effectiveness of the algorithm. The system provides a handy user interface for initializing the environment, customizing semantic types of interest, and viewing bridge concepts connecting topics.