A New Approach for Concept-Based Web Search

Abstract

Searching for information on the Internet remains a difficult task, despite considerable progress in search engines, such as Google. One difficulty for many users is to formulate a suitable search query. In this paper we propose a new interactive query refinement process that helps the user to articulate their information needs by supporting the word sense disambiguation of search terms as well as by dynamically generating potentially new relationships among several search terms, based on analysing the retrieved documents.

The main functionality of our system, called WebConceptualizer, presented in this paper support the following: 1) the user’s awareness of the different word senses of query terms. 2) to visualise a network of concepts surrounding the query terms in a graph structure that allows the user to operate at the conceptual level rather than the term level when articulating their information need. 3) the identification of documents whose content reflects a particular word sense rather than just the query words as such.

Our initial experiments with the implemented prototype shows a good accuracy in recognising the correct word sense of the main topic of the document. The user interface has not been systematically evaluated as yet. However, the usability seems rather good based on a verbal reports from a few test users.