SPECTRAL WEB CONTENT TREND ANALYSIS.

The World Wide Web as a social network reflects changes of interest in certain domains. People use the Web to present ideas, discuss thoughts and to share knowledge. It has been shown that free online content in Blogs, Wikis, News and Forums is a valuable source of information to identify trends in certain domains. The fast amount of data makes it necessary to organize the content in taxonomies or concept networks. The size of such structures does not allow experts to identify evolving aspects directly from the networks or population labels. In this paper, we present a methodology based on data mining techniques to create a domain specific concept network. We propose spectral graph analysis on the node and edge graph to identify multi conceptional and multi relational trends. We exploit the spectrum to cluster the difference graph and to highlight the changed concept network regions. The spectrum distance is further used to quantify the trend impact. Experiments on synthetic and real-world networks showed that with the proposed methodology complex multi conceptional and relational trend analysis can be performed. [ABSTRACT FROM AUTHOR]