Query term expansion and reweighting using term co-occurrence similarity and fuzzy inference

To improve the effectiveness of the classic relevance techniques for the vector model, a novel technique for term expansion and term reweighting is suggested. The advantages of the classic techniques are simplicity and good results. However, due to the simplicity, the term occurrence pattern is not considered explicitly. To supplement the classic relevance techniques, we introduce the term cooccurrence similarity as a measure of how similar the distributions within the feedbacked documents of a given term and the initial query are. With this similarity and additional information, the weight in the new query of the term is calculated by fuzzy inference. Although the experiments are performed on the small collection, the results show that the technique proposed in the paper yields substantial improvements in retrieval effectiveness.