Field-Slash Based Relevance Propagation for web information retrieval

It is evident that information resources on the World Wide Web (WWW) are growing rapidly with unpredictable rate. Under these circumstances, a major challenge in search engines is the ranking operation. Ranking is the process of ordering the returned documents in decreasing order of relevance, that is, so that the “best” answers are on the top. Over the past decade, several studies in field of relevance propagation models have been done to improve the quality of web search. These models propagate content information with the assistance of the web structure. In this paper we explore and extend one of the propagation models, called Slash-Based Relevance Propagation Model. We are going to embed a field-based similarity function in it and then conduct experimental evaluations over the extended model to know how much the extended model is effective in compared to the old one. To evaluate the extended propagation model, Letor 3.0 - a standard web test collection including two standard data sets-was used in the experiments. We have concluded that using field-based content similarity function in the propagation process provides significant improvement in the relevance propagation accuracy.