Optimizing personalized retrieval system based on web ranking

Abstract

This paper drew up a personalized recommander system model combined the text categorization with the pagerank. The document or the page was considered in two sides: the content of the document and the domain it belonged to. The features were extracted in order to form the feature vector, which would be used in computing the difference between the documents or keywords with the user's interests and the given domain. It set up the structure of four block levels in information management of a website. The link information was downloaded in the domain block level, which is the top level of the structure. In the host block level, the links were divided into two parts, the inter-link and the iutra-link. All links were setup with different weights. The stationary eigenvector of the link matrix was calculated. The final order of documents was determined by the vector distance and the eigenvector of the link matrix. © Springer-Vorlag Berlin Heidelberg 2006.