Abstract

Optimal automatic classification of web queries aid in improving the performance of search engines. To automatically classify web queries into predefined topical categories, a novel approach which transforms a classification problem to a search problem in the state space tree is proposed. The topical categories are converted to a state space tree. The directory search results and query stream are used in determining the criterion function for a node. A final traversal through the tree using best first search yields a ranked list of target categories. Experimentation on unique users of an AOL query log with previous queries yielded a comparable result to the conventional automatic web query classification techniques.

Key words: Web query classification, directory knowledge, query stream, weighted state space tree, best first search.