Document Instantiation for Relevance Feedback in the Bayesian Network Retrieval Model

Relevance Feedback consists on formulating automatically a new query, according to the relevance judgements provided by the user after evaluating the set of retrieved documents. In this paper we introduce a new relevance feedback method for the Bayesian Network Retrieval Model. This method is based on the instantiation of the observed documents as relevant or non-relevant in the Bayesian Network. We explain the theoretical bases of the model and propose different schemes for carrying out this task. The quality of the method is tested using a preliminary experimentation with different collections.