A Framework for Dynamic Topic Clustering on the Web

摘要

In this paper we present a framework for dynamic conceptual clustering of web objects (documents) based on existing clusters of users that share common interests. The basic idea is that the results of a search would be more meaningful to users when provided in the context of semantically related documents selected as "interesting" by their peers. According to the proposed approach, objects selected and ranked highly by the majority of the members of a user group form a matching object group (related to a certain topic or context). This is inherently dynamic process as new documents arrive constantly and the user groups themselves are dynamic. The paper describes formally the proposed web-search framework based on contextual topology.