Search and Ranking Algorithms for Locating Resources on the World Wide Web

Applying information retrieval techniques to the World Wide Web (WWW) environment is a unique challenge, mostly because of its hypertext/hypermedia nature and the richness of the meta-information it provides. We present four keyword-based search and ranking algorithms for locating relevant WWW pages with respect to user queries. The rst algorithm, Boolean Spread Activation, extends the notion of word occurrence in Boolean retrieval model by propagating the occurrence of a query word in a page to other pages linked to it. The second algorithm, Most-cited, is based on the number of citing hyperlinks between potentially relevant WWW pages to increase the relevance scores of the referenced pages over the referencing pages. The third algorithm, TFxIDF or vector space model, is based on word distribution statistics. The last algorithm, Vector Spread Activation, combines vector space model and spread activation model. We conducted an experiment to evaluate the retrieval eeectiveness of these algorithms. From the results of the experiment, we draw conclusions regarding the nature of the WWW environment with respect to document ranking strategies.