International Journal of applied mathematics and computer science

online read us now

Paper details

Number 4 - December 2006
Volume 16 - 2006

Guided local search for query reformulation using weight propagation

Issam Moghrabi

Abstract
A new technique for query reformulation that assesses the relevance of retrieved documents using weight propagation is proposed. The technique uses a Guided Local Search (GLS) in conjunction with the latent semantic indexing model (to semantically cluster documents together) and Lexical Matching (LM). The GLS algorithm is used to construct a minimum spanning tree that is later employed in the reformulation process. The computations done for Singular Value Decomposition (SVD), LM and the minimum spanning tree are necessary overheads that occur only initially and all subsequent work is based on them. Our experimental results reveal the effectiveness of the new technique.

Keywords
relevance feedback, clustering, latent semantic, query reformulation