Searching Biomedical Text: Towards Maximum Relevant Results

Abstract

The amount of biomedical information available to users today is large and increasing. The ability to precisely retrieve desired information is vital in order to utilize available knowledge. In this work we investigated how to improve the relevance of biomedical search results. Using the Lucene Java API we applied a series of information retrieval techniques to search in biomedical data. The techniques ranged from basic stemming and stop-word removal to more advanced methods like user relevance feedback. In our approach we developed a simple base case to be used as standard of reference. Our implemented techniques were afterwards measured up against this reference base case. We surprisingly found out that Lucene at its most basic works very well as information retrieval system in the biomedical domain. Relevance improvements beyond the base case was achieved by the introduction of user relevance feedback.