Relations between Terms Discovered by Association Rules

Abstract. This paper presents an approach to automatic knowledge base construction used for automatic query expansion. We have designed a system that is able to gain domain knowledge from analysed texts. We have used a Data Mining technique called association rules. We show how this technique is able to discover knowledge about relations between terms. We study how these relations can be used to improve retrieval performance in an Information Retrieval System (IRS). We report our experiments with the Amaryllis collections 1. Our results show that an increase in retrieval performance can be obtained using the association rules approach.