论文部分内容阅读
针对清真食品安全问题,提出一种基于句法分析的清真食品评论倾向性判别方法.首先对清真食品语料进行分析,在分词系统中添加特征词库;再利用句型分类,使用基于层叠条件随机场模型(CCRFs)的机器学习方法,将句子分为五大类(肯定句,否定句,双重否定句,反问句和疑问句)并标记;最后按句式类别对句子倾向性进行计算,并针对特殊句式进行相应处理,文中考虑了连词对评论倾向性的影响,对转折连词加以处理.实验结果表明,该方法能有效地对清真食品评论倾向性进行判别,准确率达到82.3％.“,”In view of the influence of halal food comments results of halal food safety,this paper presents a method of the tendency identification of halal food comments on the basis of syntactic parsing.First of all,it analyzes the halal food corpora and add key words in word segmentation system,and improves the accuracy of word segmentation.Secondly,the sentence classification with the cascaded conditional random fields (CCRFs) machine learning method classification based on rules,which divide the sentences into five categories:positive sentence,negative sentence,double negation,disjunctive question and interrogative sentence.Finally,it calculates the sentence tendentious by sentence classification and processes the special sentence patterns.The paper considers the influence of conjunctions on the tendency of comments,and deal with the transitional conjunctions.As a result,experiment shows that the method can effectively discriminate the tendency of halal food reviews and the accuracy is 82.3 ％.