THE ESTIMATION OF TERM RELEVANCE WEIGHTS USING RELEVANCE FEEDBACK

The term relevance weighting method has been shown to produce optimal information retrieval queries under well‐defined conditions. Unfortunately, the relevance weights cannot be determined in the absence of accurate knowledge of the occurrence frequencies of the terms in the relevant and non‐relevant documents of a collection. This study presents a realistic method for estimating the term relevance weights from information derived in an interactive search environment where relevance assessments for previously retrieved items are used later to construct improved query statements. Procedures are introduced for constructing the initial query weights by using estimated term relevance factors. These initial weights are then modified during the relevance feedback process by utilizing the occurrence frequencies of the terms in the retrieved documents obtained from an earlier search. The procedures used to construct the term relevance weights are covered in detail, and experimental output is included to illustrate the effectiveness of the methods.