Extended rough fuzzy sets for web search agent

Abstract

We propose a new method to create an intelligent Web search agent based on combining rough sets and fuzzy sets. Firstly, we create user profiles which is the important part to approximate query and document. We also design categories of user interests by grouping key terms. We update and create new categories automatically from user feedback. Secondly, we use combined rough and fuzzy sets to refine each query. Each query and document has two parts: a set of core queries and a set of possibly queries. Then we submit a set of queries to WWW search engines. Thirdly, a rough fuzzy set is used to represent each result from WWW search engines. Lastly, we compute a degree of similarity between query and documents for ranking purpose. Using these techniques, we can improve the precision level of the result from WWW search engines that are suitable for individual usage. © 2003 SRCE University Computing Centre.