Query expansion based on equi-width and equi-frequency partition

Abstract

Query Expansion has been widely used to improve the effectiveness of conceptual search. In this paper pseudo relevance feedback is used along with equi-width and equi-frequency partition technique. The proposed method effectively uses the position and frequency of the query terms for identifying a region within the retrieved documents, which is expected to contain expansion terms. This region is an intersecting region obtained by partitioning the retrieved documents using equi-width and equi-frequency partition techniques. Initial results indicate that words falling in the intersecting region contain good candidate terms for query expansion. The experiments are performed on FIRE 2011's Ad-hoc Hindi and English Data using Terrier as the retrieval engine. The initial experiments show an improvement in average precision of 12-14% in case of English data and 12.75% in case of Hindi data set.