【作者】 曹淑英；

【导师】 索红光；

【作者基本信息】 中国石油大学 ， 计算机应用技术， 2007， 硕士

【摘要】 随着Internet的普及与发展,网络已成为一个巨大的信息源。大量的数字信息在带给人们丰富便利的信息资源的同时,也给有效信息的快速获取带来了困难。自动文摘技术是自然语言处理领域的一个课题,它利用计算机自动地对文章进行处理,生成基本反映文章主题的摘要。利用自动文摘技术,人们可以减少寻找信息的时间。介绍了自动文摘技术中两个关键问题的研究。一个是文本的意义段划分问题,意义段划分是自动文摘技术研究的一个重要内容,它将文章中所讨论的不同主题划分开来,每个主题为一个意义段。在总结前人工作的基础上,从意义段划分的目的和相关的理论出发,针对传统的基于词语的TextTiling算法中信息量不足的问题,对词语进行基于HowNet的概念扩展,通过概念进行紧凑度计算。实验结果表明,基于概念扩展的TextTiling算法,在意义段划分的准确性和主题识别率上都有较大的提高。另一个是自动文摘技术中的句子计算问题,针对传统的基于词语统计的句子计算方法中缺乏语义分析的问题和基于文本结构分析的句子衡量方法中语法分析不够准确的问题,结合两种方法的思想,对词语进行基于规则的组块划分,以组块为基本单元对句子进行处理和计算。实验结果表明,基于组块的句子计算方法能够较好的提高文摘质量。最后介绍了基于用户兴趣的个性化自动文摘系统,给出了系统流程图和计算方法,通过用户对文章子主题的感兴趣程度对文摘句进行处理。实验结果表明,个性化自动文摘能够较好的满足人们的需要。更多还原

【Abstract】 With the popularization of the internet, the network has become a huge information resource. The large amounts of information not only provides us with the facilitate information resource, but also brings us a problem that how to get effective information from the internet. The automatic abstracting technology is a natural language processing topic. It automatically produces the article’s summarization which can basically reflect the article’s information and save users’searching time. Two key technologies in automatic abstracting are introduced. The first one is discourse segmentation which is an important automatic abstracting technology. It divides the article into several parts according to the subjects in the article. On the foundation of the predecessor’s works, to the lake of semantic analyzing in traditional TextTiling algorithm based on terms, concept expansion to terms based on the HowNet is taken and compact computation through the concept is made. The experimental result indicates that the TextTiling algorithm based on the concept expansion can get higher accuracy. The second one is the sentence estimation problem in automatic abstracting. By analyzing the shortage and advantage of the traditional automatic abstracting method based on terms computation and the traditional automatic abstracting method based on article grammar structure analyzing, chunks to the rule groups are built up, furthermore works of sentence processing and computation based on chunks are introduced. Experimental results show that, the summarization’s quality by the chunk computation method is improved noticeably. Finally, a system of personalized automatic abstracting is introduced. The system constructure and the computation method are showed,furthermore the summarization according to the user’s interest is produced. The experimental result indicates that, automatic abstracting based on users’ interesting analyzing can finely meet users’ needs.更多还原