Improving Information Retrieval System by Co-clustering Web Documents and Queries

Abstract

World Wide Web is considered the most valuable place for Information Retrieval and Knowledge Discovery. While retrieving information through user queries, a search engine results in a large and unmanageable collection of documents. A more efficient way to organize the documents can be a combination of clustering and ranking, where clustering can group the documents and ranking can be applied for ordering the pages within each cluster. This paper proposes an approach to co-clustering web documents and queries. When user issues a query, we construct a Query-Document Bipartite Graph from click log data. Then, we co-cluster the web documents and queries simultaneous based on the bipartite spectral graph partitioning which uses the second singular vectors of an appropriately scaled query-document matrix to yield good bipartition and rank the queries and documents on the bipartite graph via an iterative process like HITS. The results of experiments show promising improvement.