摘要
提出一种基于统计机器翻译的思想抽取评价对象与评价词的方法。该方法利用词对齐模型抽取评价对象与评价词之间的关系,并结合词共现信息等特征来估计两者关系的强度。建立一张二分图刻画评价关系,并加入领域相关性度量,利用随机游走算法迭代计算候选评价对象与评价词的置信度。在COAE2011任务3的语料上进行试验验证。结果表明,利用词对齐模型抽取评价对象与评价词可以有效提高准确度,抽取出更多的评价对象与评价词。
This paper proposed an approach to extract opinion targets and opinion words based on the idea of statistical machine translation. This method extracted the associations between opinion targets and opinion words by using word alignment model,whose strength was estimated with word co-occurrence. To model these associations,the approach constructed a bipartite graph. Then a domain relevance measure was used,and random-walk algorithm was applied to calculate the confidence of each opinion target candidates and opinion word candidates. The method was evaluated on the labeled corpus of task 3 in COAE2011. The experiment results showed that it could effectively improve the accuracy by employing the model of word-alignment to extract opinion targets and opinion words,and the method could extract more targets and words simultaneously.
出处
《山东大学学报（理学版）》
CAS
CSCD
北大核心
2016年第1期58-64,70,共8页
Journal of Shandong University(Natural Science)
基金
国家自然科学基金青年项目(61300105)
教育部博士点基金联合资助项目(2012351410010)
福建省科技重大专项项目(2013H6012)
福州市科技计划项目(2012-G-113
2013-PT-45)
关键词
评价对象抽取
评价词抽取
词对齐模型
中文句子
opinion targets extraction
opinion word extraction
word-alignment model
Chinese sentence
