一种文本分割技术的多文档文摘方法研究

ON AN MULTI-DOCUMENT ABSTRACT APPROACH BASED ON TEXT SEGMENTATION TECHNOLOGY

提出一种基于文本分割技术的多文档自动文摘方法。该方法使用HowNet作为概念获取工具,通过建立句子概念向量空间模型和利用改进的DotPlotting模型来进行文本分割。利用建立的句子概念向量空间模型计算句子重要度,并根据句子重要度、文本分割结果和文摘句相似度等因素产生文本摘要。使用ROUGE-N评测方法和F_Score作为评测指标对系统产生的文摘进行评测,结果显示使用文本分割技术进行多文档摘要是有效的。

In this paper, we propose a multi-document auto-abstract method which is based on text segmentation technology. The method uses HowNet as the conceptual access tool, and segments the text through constructing the concept vector space model （CVSM） of sentence and using improved DotPlotting model. The CVSM of sentence is used to compute the importance degree of the sentences, and the text abstract is generated according to the factors of sentences importance degree, text segmentation result and the similarity of the sentences in abstract, etc. ROUGE-N evaluation method and F_Score are used as the evaluation index to evaluate the abstract formed by the system, the results show that to use text segmentation technology for multi-document abstract is effective.