SEFIRE: A Sequential Feedback Interactive Retrieval System

The SEFIRE system incorporates pattern recognition principles in retrieving documents from hierarchically structured subgroups called category files, which are organized according to information contents and data relations. Such techniques reduce the overall search effort by efficiently eliminating at each level of the hierarchy those categories known to be unrelated to a user’s request. The hierarchical category files coupled with efficient search strategies based upon pattern recognition principles increase the overall effectiveness of interactive retrieval. This paper discusses the design principles and presents some results from our pilot study.