期刊名称:Journal of Theoretical and Applied Information Technology
印刷版ISSN:1992-8645
电子版ISSN:1817-3195
出版年度:2015
卷号:77
期号:1
出版社:Journal of Theoretical and Applied
摘要:In this paper, we present a novel approach to contrastive summarization, i.e. a specific type of summarization, which aims to compare two documents (or groups of documents) on semantic and also sentiment level. The final output of contrastive summarization is a pair of summaries, depicting what topics are most often discussed with the largest difference in opinions of the authors. We explore the possibilities of combining the latent semantic information with the information about the opinions of the authors. First, we describe related works, which show, that this problem can be approached from many different directions. Next, we present our own algorithm, based on Latent Semantic Analysis, which computes scores for excerpts of the original text and based on these, it chooses best excerpts that should be included into the final summaries. Finally, we present the evaluation of our algorithm, using speeches from Czech senate.