Proximity-based ranking of biomedical texts

Abstract

Biomedical decision making and research often require relevant evidences in the huge and ever-growing biomedical literature. Retrieval of the evidences calls for a system that accepts a natural language query for a biomedical information need, and among the large number of texts retrieved for the query, ranks relevant texts higher for access or processing. However, state-of-the-art text rankers have a weakness in dealing with biomedical queries, which often consists of several correlating concepts and prefers those texts that talk about the concepts completely. In this paper, we present a technique PRE (Proximity-based Ranker Enhancer) that measures contextual completeness of query concepts appearing in a nearby area in the text, and based on the contextual completeness, assesses the term frequency (TF) of each term in the text. Therefore, those rankers that consider TF in ranking may be supplemented with PRE, without needing to change the algorithms and development processes of the rankers. Moreover, PRE is efficient to conduct the TF assessment, and neither training process nor training data is required. Empirical evaluation shows that PRE significantly improves several state-of-the-art rankers, and is better than several state-of-the-art techniques aiming at improving rankers.