A Knowledge-Based Multi-Agent Approach for Initial Query Refinement in Information Retrieval

Query refinement is one of the main approaches for overcoming natural language lexical ambiguity and improving the quality of search results in Information Retrieval. In this paper we propose a knowledge–rich personalized approach for iterative query reformulation before sending it to search engines and entropybased approach for refinement quality estimation. The underling hypothesis is that the query reformulation entropy is a valuable characteristic of the refinement quality. We use multi-agent architecture to implement our approach. Experimental results confirm that there is a trend for significant improvement in the quality of search results for large values of entropy.