计算机工程 ›› 2012, Vol. 38 ›› Issue (16): 170-173.doi: 10.3969/j.issn.1000-3428.2012.16.044

• 人工智能及识别技术 • 上一篇    下一篇

基于指代消解和篇章结构分析的自动摘录算法

郑 诚,刘福君,李 清   

  1. (安徽大学计算机科学与技术学院,合肥 230039)
  • 收稿日期:2011-10-20 修回日期:2011-12-05 出版日期:2012-08-20 发布日期:2012-08-17
  • 作者简介:郑 诚(1966-),男,副教授、博士,主研方向:语义信息检索,数据挖掘;刘福君、李 清,硕士研究生
  • 基金项目:
    安徽省自然科学基金资助项目(11040606M133)

Automatic Extraction Algorithm Based on Anaphora Resolution and Text Structure Analysis

ZHENG Cheng, LIU Fu-jun, LI Qing   

  1. (School of Computer Science and Technology, Anhui University, Hefei 230039, China)
  • Received:2011-10-20 Revised:2011-12-05 Online:2012-08-20 Published:2012-08-17

摘要: 传统自动文摘方法生成的文摘结果指代关系模糊,且对于某些段落结构有规律的文章,没有分析文章结构与主题思想之间的关系。为此,提出一种基于指代消解和篇章结构分析的自动摘录算法。采用有限知识的思路完成指代消解,利用指代消解解决文摘语义不连贯问题,以提高句子权重计算的准确性,对文章做主题划分时进行篇章结构识别,按照段落标题信息划分段落结构。实验结果表明,该算法在受限金融领域文本自动摘录中,具有较高的准确率和召回率。

关键词: 自然语言处理, 自动摘录, 向量空间模型, 主题划分, 篇章结构, 指代消解

Abstract: There are some problems should be considered in automatic extraction of traditional methods: Conference relations in the result of automatic extraction are not clear, some relationships between obvious structures of paragraphs and the theme of the text are not paid enough attention. For which, this paper presents a method based on anaphora resolution and text structure analysis, which combines the traditional statistics with regulars on automatic abstract. This method applies limited knowledge to pronoun resolution, which is to solve the problem of semantic incoherence, also to improve the precision when computing sentences’ weight. Based on sequential paragraphic similarity, this method can recognize obvious topics to partition text. Experimental results show that this method improves precision and recall when it is applied for limited-financial field.

Key words: Natural Language Processing(NLP), automatic extraction, Vector Space Model(VSM), topic segmentation, text structure, anaphora resolution

中图分类号: