计算机与现代化

• 数据库与数据挖掘 • 上一篇    下一篇

基于综合支持度的广义空间实体关系特征词提取方法

  

  1. (国防科学技术大学电子科学与工程学院,湖南长沙410073)
  • 收稿日期:2014-05-12 出版日期:2014-08-15 发布日期:2014-08-19
  • 作者简介:姜伟(1989-),男,江西上饶人,国防科学技术大学电子科学与工程学院硕士研究生,研究方向:摄影测量与遥感; 钟志农 (1975-),男,副教授, 博士,研究方向:数据挖掘,地理信息处理; 吴烨(1986-),男,博士研究生,研究方向:信息系统与信息技术。

Method of Abstracting Feature Words of Generalized Spatial Entity Relationship #br# Based on Synthesis Supporting Probability

  1. (College of Electronic Science and Engineering, National University of Defense Technology, Changsha 410073, China)
  • Received:2014-05-12 Online:2014-08-15 Published:2014-08-19

摘要:

广义空间实体关系特征词能够正确表示两实体之间的关系及其语义信息。为了准确地提取关系特征词,基于关系候选词的位置、
词性和词频等特征,定义综合支持度,用于评价该词成为关系特征词的可能性。基于综合支持度,提出关系特征词提取模型。最后对搜
狐网页新闻中的组织机构与人物(ORG-PER)和地名与人物(LOC-PER)两类数据进行测试,实验表明,该模型提取关系特征词的准确率
达到90%左右。

关键词: 广义空间实体, 关系提取, 关系特征词, 候选词, 支持度

Abstract:

 Feature words of relation between generalized spatial entities can express these relations and semantics
of them accurately. In order to abstract feature words precisely, firstly, based on location, speech and frequency
of candidate word around entities, this paper defined synthesis supporting probability, aiming at evaluating the
likelihood of becoming feature word. Secondly, a mode for abstracting feature words was proposed. Finally,
experiments were carried on ORG-PER data and LOC-PER data in Sohu news, which achieved a precision about 90%.

Key words: generalized spatial entity, relation abstracting, feature word, candidate word, supporting probability

中图分类号: