An Ensemble Approach for Expanding Queries
Abstract
In our TREC participation, we used an ensemble approach in query expansion. Query expansion, such as synonym expansion, had shown promising results in medical literature search. On the other hand, some of the 2011 papers reported worse results from expansion. Since there are multiple knowledge sources available and each resource has clear strengths and weaknesses, we tested the combination of three expansion methods versus each individual method. We found that the ensemble approach performed better (in terms of average infAP, infNDCG, R-prec, and P10) than the individual methods and better than the Lucene baseline. The individual expansion methods, however, did not improve the baseline Lucene performance. We also performed an unofficial run using a concept index to boost the query performance, which led to small improvements in infAP, infNDCG, and R-prec.
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 01, 2012
- Accession Number
- ADA579309
Entities
People
- Doug Redd
- Duy Bui
- Qing Zeng-treitler
- Thomas Rindflesch
Organizations
- University of Utah