Abstract:

PageRank is a popular query independent link-analysis algorithm for ranking web pages. The Intelligent Surfer algorithm is one of proposals describing how to combine PageRank and user query context for search engine results personalization. The cost of this solution is considerable increase of the computation time. Efficiency and speed of the PageRank computation is very important due to the enormous size of the ever growing Web’s hyperlink structure. This paper describes CZDIS algorithm - The Intelligent surfer algorithm improvement proposal when using Czech language model to reduce the number of considered terms. This leads to better efficiency of computation and ranking. Further more, proposal of paralell solution of sparse linear system, instead of traditional eigenvector problem, is included. Distributed computation leads to the massive reduction of computation time. Experiments using large subset of Web pages in Czech language prove that this algorithm significantly outperforms Intelligent surfer in the time of computation, while conserving its quality personalized ranking of the returned pages.

Keywords:

web ranking, page rank, intelligent surfer, distributed computing.