Abstract—This paper focuses attention on summarizing news articles using graph based approaches. The foundation for the graphical techniques is the adjacency matrix evaluated based on a suitable similarity measure between the sentences of the document. Two techniques, cumulative sum proposed by us and the degree of centrality method proposed by Erkan etal. are investigated. We also propose a recursive method of repeatedly using the above two methods after discounting the already selected sentences. We introduce two metric Effectiveness1 and Effectiveness2 to evaluate the summaries prepared by the system in comparison to the ‘golden standard’ summary prepared by the human judges. Comprehensive investigations with single and multi document summaries show that the discounting methods are superior to their basic counterparts and provide promise and scope for further improvements.
Index Terms—Single/multi document summarization; similarity measure; degree centrality; Effectiveness; evaluation
Shanmuasundaram Hariharan is with Department of Information Technolgy, B. S. Abdur Rahman University, Chennai, Tamilnadu, India. (Phone: 04422751347, Mobile: 9884204036, He is working as Assistant Professor and currently pursuing his doctoral studies in the area of Information Retrieval.
Rengaramanujam Srinivasan is with B. S. Abdur Rahman University, Chennai, Tamilnadu, India. Currently he is working as Professor in Department of Computer Science and Engineering)
Cite: Shanmugasundaram Hariharan and Rengaramanujam Srinivasan, "Studies on Graph based Approaches for Singleand Multi Document Summarizations," International Journal of Computer Theory and Engineering vol. 1, no. 5, pp. 519-526, 2009.
Copyright © 2008-2024. International Association of Computer Science and Information Technology. All rights reserved.
