摘要： 通过对自动文摘技术的研究，针对叙事类文本，以事件作为基本语义单元，提出一种基于事件的多主题文本自动文摘方法。利用事件和事件间的关系构建事件网络文本表示模型，使用社区划分算法解决子事件主题划分问题。实验结果表明，该方法提取出的摘要准确率、召回率及F值较高，能更好地概括文本的内容。

关键词: 多主题文本, 自动文摘, 文本表示, 图结构, 事件网络, 主题划分

Abstract: By studying the technology of automatic summarization, this paper considers the event as a basic semantic unit for narrative texts, and presents a new event-based method for automatic summarization in multiple topics text. This method utilizes events and the relationship between events to build event-network text representation model, and uses the algorithm of community detection to solve the problem of sub-event topic partition. Experimental results show that this method extracts a summary of higher precision, recall and F vaule, and better summarizes the content of the text.

Key words: multiple topics text, automatic summarization, text representation, graph structure, event network, topic partition

中图分类号: 

• TP311