FDUMedSearch at TREC 2015 Clinical Decision Support Track

Dongfang Road, Shanghai, 200127, China, ! Abstract.This paper describes the participation of the FDUMedSearch team at TREC 2015 Clinical Decision Support track (CDS2015). Given the medical cases, the main purpose of CDS2015 is to develop effective information retrieval techniques in finding relevant documents for patient care. We used Indri as the retrieval engine, which implemented query likelihood method as the baseline. In addition, query expansion using Medical Subject Headings (MeSH), pseudo relevance feedback and classification were used to enhance the retrieval performance. We also tried to extract keywords in two different ways, automatically and manually. Experimental results show that our method achieved significant improvement over baseline methods. !