摘要
介绍了一种基于知识库的方式获取文本中讨论线索的方法。该方法根据文本出现的控制词与概念网中各个概念的关联,找出文本以及文本中各个段落的主题概念,并通过对概念网中的概念进行关系计算,获取文本的讨论线索,较好地避免了因文本叙述方式的不同以及指代的存在而带来的影响.实验结果显示,该方法对文本主题识别的准确率达82%,对段落主题识别的准确率达70%.
A repository-based method to acquire the discussion clues in the text is presented in this paper. In this method, the topic concepts of the text and each paragraph are found out on the basis of the relevance between the control words in the text and each concept in the concept net, and the discussion clues are acquired on the basis of the relevance computation of each concept. Thus, the effects of different depiction manners and the existing replacements are avoided. Experimental results show that, by using the proposed method, the accuracy of text topic identification reaches 82% and that of paragraph topic identification reaches 70% .
出处
《华南理工大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2004年第z1期96-98,共3页
Journal of South China University of Technology(Natural Science Edition)
关键词
文本
讨论线索
自动获取
自动文摘
text
discussion clue
automatic acquirement
automatic summarization