Recently web pages genre identification attracts more attentions because of its importance in web searching. Most of existing works used the features extracted from web pages and applied machine learning approaches like SVM as classifier to identify the genre of web pages. However, in the case where web pages do not contain enough information, such an approach may not work well. In this paper, we consider to tackle genre identification in such situations. We propose a link-based graph model that taking into account neighboring pages but greatly reducing the noisy information by selecting an appropriate subset of neighboring pages. We evaluated this neighboring pages based classifier with other classifiers. The experiments conducted on two known corpora, and the favorable results indicated that our proposed approach is feasible.
展开▼
