Leveraging Search Engine Results for Query Classification

Full Paper (13 pages)
PDF (195KB)
Shruti K. Bhandari and Brian D. Davison

Web query classification is significant to search engines for the purpose of efficient retrieval of appropriate results in response to user queries. User queries are short in nature, contain noise and are ambiguous in terms of user intent. In this paper, we present different features--such as snippets, page content and titles of search engine results for a given query--that can be used to enhance user queries for classification purposes. We train various classifiers using different features so that these classifiers can be evaluated on a set of test queries. We show that the performance of our technique of classification of a query using the snippets of search engine results for that query is comparable to that obtained by the solutions provided by the winning teams at the KDD Cup 2005 competition, in spite of the fact that our technique is less complex in comparison to the other winning solutions. It also noticeably outperforms query classification using query text directly or the content of pages returned by search engines for the query.

Technical Report LU-CSE-07-013, Dept. of Computer Science and Engineering, Lehigh University, 2007.

Back to Brian Davison's publications

Last modified: 28 May 2011 Brian D. Davison