Abstract: In order to solve the greatly dimension of word frequency and sparse matrix, this paper proposes a multi-document summarization method based on sub-topics area partition. It uses HowNet to obtain concept of word, constructs Concept Vector Sapce Model(CVSM) replace traditional Word Frequency Vector Space Model(WFVSM). After constructed CVSM, the document is segmented into several units in terms of the sub-topics in the document. The most representative sentences in each sub-topic unit are selected as the summary sentences. Experiment results prove the validity of the method.

Key words: sub-topic area, automatic summarization, HowNet, Concept Vector Space Model(CVSM)

摘要： 为解决词频矩阵的词频维数过大和矩阵过于稀疏的问题，提出一种子主题区域划分的多文档自动文摘方法。使用知网进行概念获取，建立概念向量空间模型，代替传统的词频向量空间模型。在概念向量空间模型的基础上，利用一种改进的层次分割法对文档集合进行子主题划分，从各个子主题中抽取出满足一定数量的句子作为文摘。实验结果验证了该方法的有效性。

关键词: 子主题区域, 自动文摘, 知网, 概念向量空间模型

CLC Number: 

• TP391