Extracting Multi-document Summarization Based on Local Topics

In this paper, we propose a new method for text summarization. The system finds topic word and event word firstly, and then recalculates word weight. Using recalculated word weight to compute similarly of paragraphs to search local topics units. The most representative sentences in each local topic unit are selected as the summary sentences. By analyzing semantic structure of the documents first, the summary sentences are not redundancy and the coverage of each local topic is balanced. Experimental results show that our approach is effective and efficient, and performance of the system is reliable.