Topic Hierarchy Generation Based on Query Expansion and Anchor Text with Term-correlation

Abstract

With the accelerated growth of the Internet, a ”Topic Directory Query” service becomes increasingly important to help users easily find information of interest by choosing the right topics during their search. This paper proposes a novel scheme that employs query expansion, anchor text and term-correlation to construct and generate topic hierarchy automatically. Query expansion is first adopted to expand the scope of a term through the powerful meta-search engine. The anchor text of meta-search results is then applied, saving significant computational efforts. Finally, topics are grouped by appropriate term-correlation computations. Results of experiments focusing on computer science terms demonstrate that the proposed approach outperforms systems with manually constructed topic hierarchies, e.g., the Open Directory Project and Yahoo Web Directory, especially searching in topics that are not manually constructed in advance by those systems.