Supporting relevance feedback with concept learning for semantic information retrieval in large OWL knowledge base

Abstract

Relevance feedback in information retrieval is a popular way to learn the user’s intents. We investigate the feasibility and methodology of applying the concept learning in OWL knowledge base to deal with feedback in interactive information retrieval system. The feedback from the initial search results is considered as examples, and then the inductive concept learning technique is employed to generate a concept that describes the user’s requirement. We deal with the performance of concept learning by reducing the scale of the problem; a clustering based OWL knowledge base partitioning method is proposed to divide the knowledge base into several small-scale ones with acceptable recall and precision. An interactive healthcare information retrieval prototype is developed for evaluation; the results of the user study and precision-recall graph show the efficiency of the methods proposed.