Categorizing Queries and Expanding Keywords with a Coreference Graph

摘要

A method and apparatus is provided for determining related keywords to narrow a query, and/or for categorizing a query. A keyword graph connects keyword nodes to each other based on degrees of cross-reference indicating how frequently keywords associated with the nodes appear in searches. A domain node representing a category hooks to a category-matching node in the keyword graph. Based at least in part on a degree of cross-reference between another node and the category-matching node, the domain node hooks to the other node. Alternately, the domain node hooks to nodes that match user-identified keywords. At query time, the query is categorized by the domain node closest to a node matching the query. Keywords related to the category may be determined from the nodes that are hooked to the domain node. The related keywords can be used to narrow a search or expand the metadata of a document.