A Significance-Driven Framework for Characterizing and Finding Evolving Patterns of News Networks

Leiming Yan, Jinwei Wang, Jin Han, Yuxiang Wang

Abstract: Social network analysis has become extremely popular in recent years. What are the most significant evolving behaviors in a social network? It is very difficult to find significant evolving behaviors from a large network in a long evolving time interval. Besides, verifying and evaluating enormous dynamic patterns extracted from a large social network by experts are also too hard to generalize well. In this work, a significance-driven framework is proposed to characterize the evolution of local topology and find dynamic patterns with evidently statistical significance for temporally varying news report networks. Two significance indices—potential index and evolving score are introduced for evaluating evolving patterns. Finally, we present a systematic analysis of one real news network, which demonstrates that the method we proposed can find the evolving characteristic and extract significant dynamic patterns from news networks.