AUEB at TREC 2015: Clinical Decision Support Track

One of the goals of clinical decision support systems is to provide physicians information about how to best care for their patients. The Clinical Decision Support track organized by TREC, focuses on developing new techniques to retrieve articles from the biomedical literature relevant to the medical records of the patients. Due to the large volume of the existing literature and the diversity in the biomedical field, this is a very challenging task. This paper describes the two medical information retrieval systems designed by the Athens University of Economics and Business for participation in the 2015 Clinical Decision Support track. The two systems share many common features. Both made use of bigrams along with unigrams for repesenting the documents. Both systems performed automatic query expansion using popular medical knowledge bases. However, the two systems employed different strategies to index the corpus which led to different retrieval methods. One utilized the vector space model with tf − idf term weighting, while the other the vector space model with tw − idf term weighting. The results showed that tf − idf outperformed tw − idf.