基于最大熵与Bootstrapping的关联三元组识别方法

基于〈产品特征，情感词〉关联对的缺点，讨论了情感词与否定性副词搭配的必要性，提出了〈Pfeature，FIag，sword〉关联三元组，能够更准确地表达文本中相关评论句对产品特征的情感倾向。采用两个步骤来提取关联三元组：首先，利用已训练好的最大熵模型作为分类器，结合Bootstrapping方法完成了产品特征与情感词语关联对的抽取；其次，将情感词前的否定性副词抽取出来，合成关联三元组。

Correlative Triple Recognition based on Maximum Entropy and Bootstrapping

Based on the shortcoming of ＂Product feature, Sentimental word＂, the new concept of ＂Pfeature, Flag, Sword＂ arc, proposed after discussing the necessary to match the sentimental words with denying adverbial words,which can precisely show the objective sentimental tendency of the text sentence. Two steps are adopted to acquire the Correlative Triple. First, the pair〈Pfeature, Sword〉 is gained by combining Bootstrapping and the Maximum Entropy model trained well as a classifier. Second, the negative adverbs before sentimental words are tricked up by using an alzorithm, they consist of Correlative Triple.