Skip to Main Content
In this paper, we propose a unified retrieval model that is called the hyper-textual language model for Web information retrieval. The proposed model seamlessly integrates information from multiple sources including Web content, URL, hyperlinks, and the topology of the Web in a unified modeling framework. On the one hand, this model extends the language modeling technique to accommodate special structural and semantic information brought by the hyperlinks of the Web; on the other hand, it provides a formal retrieval model that realizes topic-relevant pageranking. Preliminary experimental study on a university website shows that the performance of this formal retrieval model is comparable with the performance of the Googlepsilas University Search.