图书情报工作 ›› 2011, Vol. 55 ›› Issue (06): 57-60.

• 情报研究 • 上一篇    下一篇

基于汉语框架网的语义角色标注算法

赵文娟1,闫红梅2,王蔚林2   

  1. 1. 山西大学商务学院信息学院
    2. 山西大学人事处
  • 收稿日期:2010-09-20 修回日期:2010-11-08 出版日期:2011-03-20 发布日期:2011-03-20
  • 通讯作者: 赵文娟

Semantic Role Labeling Algorithm Based on Chinese FrameNet

Zhao Wenjuan 1,Yan Hongmei 2,Wang Weilin 2   

  1. 1. School of Information, Business College of Shanxi University,
    2. Staff Department, Shanxi University,
  • Received:2010-09-20 Revised:2010-11-08 Online:2011-03-20 Published:2011-03-20
  • Contact: Zhao Wenjuan

摘要: 在汉语框架网(CFN)的基础上,介绍语义角色自动标注的步骤和流程,提出基于文本匹配和最大熵分类器的语义角色自动标注方法。在文本匹配算法中,综合考虑短语类型、短语相对于目标词位置、句法功能三个因素及其对句子相似度影响的权重;在最大熵算法中,也尝试一些新的特征及其组合,最后利用例子对该方法进行有效性验证。

关键词: CFN, 汉语框架网, 语义角色, 词性匹配, 最大熵, 自动标注

Abstract: Based on Chinese FrameNat(CFN), the paper introduces the automatic annotation steps and process of semantic roles, and gives an automatic annotation method based on text matching & maximum entropy annotation. In text matching algorithm, the paper comprehensively considers the phrase type, phrase relative to target word position, syntactic function, and the three factors of sentence similarity affecting weight. In maximum entropy algorithm, the paper tries some new features and their combinations. Finally, the paper uses an example to test the effectiveness of this method.

Key words: CFN, Chinese FrameNet, semantic role, POS matching, maximum entropy, automatic labeling