Hidden Markov Model Based Web Query Classification

Free registration required

Executive Summary

Topical classification of web queries aids in the efficient retrieval of information and in profitable online advertising. In this paper, four-tier architecture is proposed for the automatic classification of web queries into topical categories; directory knowledge and query log generate the feature set for further processing, a glossary based classifier to produce an unordered set of pre-defined topical categories, a HMM to incorporate returned search results and prior knowledge of the user into the architecture and estimates the maximum likelihood of a sequence of topical categories using the Viterbi decoding algorithm.

  • Format: PDF
  • Size: 126.94 KB