Vector space based on hierarchical weighting: a component ranking approach to component retrieval

In this paper, we present an approach to software component ranking intended for use in searching for such components on the internet. The method used introduces a novel method of weighting keywords that takes account of where within the structure of a component the keyword is found. This hierarchical weighting scheme is used in two ranking algorithms: one using summed weights, the other using a vector space model. Experimental comparisons with algorithms using TF-IDF weighting that ignore component structure are described. The results demonstrate consistent superiority of the hierarchical weighting approach.