TAX-PQ: dynamic taxonomy probing and query modification for topic-focused Web search

We propose a novel Web search scheme TAX-PQ. TAX-PQ enables taxonomy-based topic-focused Web search on ordinary Boolean Web search interfaces. TAX-PQ utilizes a taxonomy and the data set maintained in an existing taxonomy-based search facility for this purpose. The search is initiated by designating an initial query and a context category in the taxonomy. The data set in the taxonomy-based search facility is probed with a technique combining the initial query with sampling, and a decision tree is constructed from the sampled query result. A query modifier is then derived from the decision tree to focus the initial query on the selected context category. To adapt TAX-PQ to different user requirements on search result effectiveness and properties of target Web search interfaces, we have developed a new decision tree construction algorithm. Experiments involving real Web sites show that TAX-PQ can significantly improve the Web search process and result. The results comply with user requirements under constraints of the target Web search interfaces.