Applying Cross-Topic Relationships to Incremental Relevance Feedback

General purpose search engines like Google and Yahoo define search topics for the purpose of document organization, yet their hierarchical structures cover only a portion of topic relationships. Search effectiveness can be improved by using search topic networks, in which topics are linked through semantic relations. In our search model, is-child and is-neighbor relations are defined as relations among search topics, which in turn can serve as search techniques; the is-child relation allows searching from general concepts, while the is-neighbor relation provides fresh information that can help users to identify search areas. This search model uses the Bayesian Networks and the incremental relevance feedback. Our experiments show that search models using the Bayesian Networks and the incremental relevance feedback improve search effectiveness.