Geodesic distance based multi-document summarization

Huifang Ma, Qing He, Zhongzhi Shi

Abstract: This paper presents a novel extractive approach which takes advantage of geodesic distance for sentence similarity computation to multi-document summarization task. Based on geodesic distance between every two sentences, the text relationship map is constructed. Sentences with higher degree in the map are selected and grouped into clusters. Finally, sentences with highest degree of each cluster are chosen to form the final summary. Experiments on DUC data are performed and the ROUGE evaluation results demonstrate that the proposed approach can be an effective way for multi-document summarization.