Using Probabilistic Argumentation Systems to Search and Classify Web Sites

As the amount of information stored on the web increases at an amazing pace, it gets harder for search engines to retrieve the needed information. Recently, attention in the web community has focused on the use of hyperlinks to help index and organize information. Several methods have been suggested to take account of the knowledge induced by hyperlinks, with different objectives in mind. In this paper, we focus on three of them: improving document ranking, estimating the popularity of a web page, and extracting the most important hubs and authorities related to a given topic. Using probabilistic argumentation systems, a technique for dealing with uncertain knowledge which integrates propositional logic and probability theory, we show how all these techniques can be modeled in a unified logical framework. This allows comparisons of the different methods for using hyperlinks and illustrates some of their weaknesses based on some experiments.