Neural networks-based multi-interest information filtering

With an increasing amount of information available, users find it difficult to obtain the most relevant Web pages from the large number returned by search engines. This is particularly true if users have more than one interest and require information spanning several of them. In this paper, we present an improved adaptive neural fuzzy network providing an information filtering system for the Web to sift the results provided by external search engines. We discuss how to model a user's multi-interests and filter information according to 'IF-THEN' rules and how to optimize and adjust the parameters stored in the network as a user's multi-interests. Preliminary experiments show that our prototype system improves the performance of current search engines for the user who has multi-interests. The distinguishing features are that of a user model embedded in a neural fuzzy network to process the multi-interests and the replacement of the traditional cosine measure method by a parameterized non-linear map allowing multi-interests to be processed. The results achieved support the choice of the neuro-fuzzy network for multi-interest information filtering and show that the information retrieval system can benefit from the technology available in Soft Computing for better retrieval.