PostingRank: Bringing Order to Web Forum Postings

Abstract

Web forum is an important information resource. Each day innumerable postings on various topics are created in thousands of web forums in internet. However, only a small part of them are utilized for the reason that it is difficult to rank the postings importance. Unlike general web sites with hyperlinks in web pages created by editors, links in web forums are automatically generated, therefore, traditional link-based methods, such as famous PageRank are useless to rank postings. In this paper, we propose a novel algorithm named PostingRank to rank postings. The main idea of our method is to exploit the common repliers between postings and leverage the relationship between these common repliers. We build implicit links based on that co-repliers relation and construct a link graph. In the way of iterative calculation, each posting’s importance score can be obtained. The experimental results demonstrate that our method can improve retrieval performance and outperforms traditional link-based methods.

This work is sponsored by National Basic Research Program of China Project NO.2002CB312006, National Natural Science Funds of China Project NO.60673008 and National High Technology Research and Development Program of China Project NO.2007AA04Z135.