Adaptive Combinational Web Ranking Algorithm Using Coarse-grained and Fine-grained Features

The proliferation and huge size of information in the web is an important issue in response to the user query in information retrieval. Currently, different ranking algorithms such as PageRank and BM25 have been proposed. In this paper an adaptive combinational ranking algorithm for achieving to higher precision and efficiency is proposed. This algorithm uses combination of Coarse-grained features such as BM25 and TF-IDF and also fine-grained features such as Term Frequency and In-Degree of pages, in the learning process to produce better results. In learning process, the Ordered Weighted Aggregation Operator (OWA) operation and experts judgments are used. For evaluation we used a standard test collection called LETOR including “WEB TREC 2004” and we found considerable results.