In the last few years, a lot of research has been devoted to developing new techniques for improving the recall and the precision of current web search engines. Few works deal with the interesting problem of identifying the communities to which pages belong. Most of the previous approaches try to cluster data by means of spectral techniques or by means of traditional hierarchical algorithms. The main problem with these techniques is that they ignore the relevant fact that web communities are social networks with distinctive statistical properties. In this paper we analyze web communities on the basis of the evolution of an initial set of hubs and authoritative pages. The evolution law captures the behaviour of page authors with respect to the popularity of existing pages for the topics of interest. Assuming such a model, we have found interesting properties of web communities. On the basis of these properties we have proposed a technique for computing relevant properties for specific topics. Several experiments confirmed the validity of both the model and identification method.
Web communities: Models and algorithms
GRECO, Gianluigi;GRECO, Sergio;ZUMPANO, Ester
2004-01-01
Abstract
In the last few years, a lot of research has been devoted to developing new techniques for improving the recall and the precision of current web search engines. Few works deal with the interesting problem of identifying the communities to which pages belong. Most of the previous approaches try to cluster data by means of spectral techniques or by means of traditional hierarchical algorithms. The main problem with these techniques is that they ignore the relevant fact that web communities are social networks with distinctive statistical properties. In this paper we analyze web communities on the basis of the evolution of an initial set of hubs and authoritative pages. The evolution law captures the behaviour of page authors with respect to the popularity of existing pages for the topics of interest. Assuming such a model, we have found interesting properties of web communities. On the basis of these properties we have proposed a technique for computing relevant properties for specific topics. Several experiments confirmed the validity of both the model and identification method.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.