针对微博观点句识别及情感极性分类任务,提出了基于增强字向量的微博观点句情感极性分类方法.使用单字作为句子表示的基本单元,同时在单字中嵌入了该字所在的词信息以及该词的词性信息,以此训练得到字向量替代传统的词向量融合的句子表示方法.使用基于K-means的方法对向量化的句子进行情感判别,仅需要对文本进行分词和词性标注,无需额外的语言学资源.在COAE2015任务2的微博句子数据集上进行测试,取得了较好的结果.%A method based on enhanced-character vector was proposed for the identification and classification of Weibo opinion sentences.The characters, which embedded the corresponding information of words and part-of-speech, were used as the basic unit for sentence representation instead of words.Then a method based on K-means was adopted for sentence polarity classification.The proposed method was almost free of linguistic resources (except POS tags), and required no elaborated linguistic rules.Experiment on the data of COAE2015 task 2 showed that this method achieved good results.
展开▼
