Efthimis N. Efthimiadis^1, David G. Hendry^1, and Chong-Ki Tsang^2

1 Information School, University of Washington
2 Microsoft

Experiments in Query Expansion

UW/Microsoft Symposium, 2/16/07

In this study we evaluate different QE approaches using the TREC collection. We employed the half collection evaluation methodology and evaluated three relevance feedback approaches (pseudo-relevance feedback, simulated user feedback, random feedback), multiple term selection methods and QE combinations. We further compared the effect of stemming (Porter vs. Krovetz vs. not stemming) and the use of stop lists vs. non-stopping to retrieval effectiveness.

Back to symposium main page