A user-adaptive neural network supporting a rule-based relevance feedback

Associative mechanisms, such as those based on the use of thesauri, document clustering and relevance feedback, are widely employed in information retrieval systems to enhance their effectiveness. They make it possible to retrieve also the documents not directly indexed by the search terms. In this paper, a relevance feedback model is defined, based on an associative neural network in which concepts meaningful to the user are accumulated at retrieval time by an iterative process. The network can be regarded as a kind of personal thesaurus of the user. A rule-based superstructure is then defined to expand the query evaluation with the meaningful terms identified in the network. The search terms are expanded by taking into account their associations with the meaningful terms in the network.