Experiments on Genomics Ad Hoc Retrieval

This paper represents te results of the State University of New York at Buffalo (School of Informatics) in the TREC 2005 Conference. We participated in the Genomics ad hoc retrieval task. Our approach used the SMART system for indexing the large collection of MEDLINE documents. For this purpose we used a distributed retrieval approach and divided the large collection into 5 non overlapping sub collections. We tried several approaches on the training topics to select the best run possible. Our results perform slightly above the median system in the conference. We also paired with the NLM team to contribute a run for their fusion approach.