【作者】 王宏显； 周强；

【Author】 Wang Hongxian Zhou Qiang Center for Speech and Language Technologies,Research Institute of Information Technology,Tsinghua University Center for Speech and Language Technologies,Division of Technology Innovation and Development,Tsinghua National Laboratory for Information Science and Technology Department of Computer Science and Technology,Tsinhgua University,Beijing 100084,China

【机构】 清华大学信息技术研究院语音和语言技术中心 清华信息科学技术国家实验室技术创新与开发部语音和语言技术中心； 清华大学计算机科学与技术系；

【摘要】 特征选取是用统计方法进行词义排歧的关键。本文通过句法块的分析结果,将主语和宾语、以及主语和宾语在知网中的归类信息应用于目标动词的词义排歧。实验表明,句法特征对于目标动词的词义排歧有重要作用,相对于仅使用词语和词性特征的系统,加入人工标注句法特征后,拥有类动词正确率由83.7%提高到89.2%,存在类动词正确率由84.8%提高到86.1%。更多还原

【Abstract】 Feature selection is the key step in statistical method based word sense disambiguation.Based on the result of chunk parsing,we apply the subject,object and their categories in HowNet to target verb sense disambiguation.The experimental result shows that,the syntactic feature is useful for target verb sense disambiguation.Compared with using word and POS feature only,the accuracy increases from 83.7%to 89.2%after the syntactic feature added.更多还原