Sub-Topic Segmentation in Multi-Document

Abstract

The similar sentences in multi-document set are combined into one class, and each class is one sub-topic. Describing the sub-topics from the perspective of understanding makes the multi-document summarization become the one with greater coverage and less redundancy. This paper presents a sub-topic segmentation method based on maximum tree algorithm. And based on sentences similarity matrix, maximum tree is calculated, as well as the sub-topic segmentation is realized through the analysis of the different communities for the sub-topic. The experiment shows that the method achieves the desired result.