Using Semantic-Based Association Rule Mining for Improving Clinical Text Retrieval

Abstract

AbstractAssociation rule (AR) mining has been widely used on the electronic medical records (EMR) for discovering hidden knowledge and medical patterns and also for improving the information retrieval performance via query expansion. A major obstacle in association rule mining is that often a huge number of rules are generated even with very reasonable support and confidence. The main challenge of using AR in information retrieval (IR) is to select the rules that are related to the query, since many of them are trivial, redundant or semantically wrong. In this paper, we propose a novel approach to modeling medical query contexts based on mining semantic-based AR for improving clinical text retrieval. We semantically index the EMR with concepts of UMLS ontology. First, the concepts in the query context are derived from the rules that cover the query and then weighted according to their semantic relatedness to the query concepts. The query context is then exploited to re-rank patients records for improving clinical retrieval performance. We evaluate our approach on the medical TREC dataset. Results show that our proposed approach allows performing better retrieval performance than the probabilistic BM25 model.