Correlating summarization of multi-source news with k-way graph bi-clustering

  • Authors:

  • Ya Zhang;Chao-Hsien Chu;Xiang Ji;Hongyuan Zha

  • Affiliations:

  • The Pennsylvania State University, PA;The Pennsylvania State University, PA;NEC Laboratories America, Cupertino, CA;The Pennsylvania State University, PA

  • Venue:
  • ACM SIGKDD Explorations Newsletter
  • Year:
  • 2004

Quantified Score

Hi-index 0.01

Visualization

Abstract

With the emergence of enormous amount of online news, it is desirable to construct text mining methods that can extract, compare and highlight similarities of them. In this paper, we explore the research issue and methodology of correlated summarization for a pair of news articles. The algorithm aligns the (sub)topics of the two news articles and summarizes their correlation by sentence extraction. A pair of news articles are modelled with a weighted bipartite graph. A mutual reinforcement principle is applied to identify a dense subgraph of the weighted bipartite graph. Sentences corresponding to the subgraph are correlated well in textual content and convey the dominant shared topic of the pair of news articles. As a further enhancement for lengthy articles, a k-way bi-clustering algorithm can first be used to partition the bipartite graph into several clusters, each containing sentences from the two news reports. These clusters correspond to shared subtopics, and the above mutual reinforcement principle can then be applied to extract topic sentences within each subtopic group.