ECNU at TREC 2016: Web-based query expansion and experts diagnosis in Medical Information Retrieval

In this paper, we present our work in TREC 2016 Clinical Decision Support Track. Among five submitted runs, two of them are based on summary topics and the others on note topics. In summary version run, we expand the original text with external data on web. Note topics are much longer than the summary, which contain a significant number of medical abbreviations as well as other linguistic jargon and style. An automatic method and a manual method are applied to process note topics. In the automatic method, we utilize KODA, a well-known knowledge drive annotator, to extract key information from the original text. In the manual one, we ask medical experts to diagnose and give their advice. For all of the five runs, we adopt Terrier search engine to implement various retrieval models. Furthermore, results combinations are applied to improve the performance of our model.