基于词向量和情感本体的短文本情感分类

Short text sentiment classification:Based on the word vector and emotional ontology

目前短文本情感分类主要采取统计自然语言处理、情感语义特性两种方式,而将这两种方式相结合进行情感分类的研究较少,故提出将这两种方式进行结合,设计基于词向量与情感本体相融合的短文本情感分类方法。首先利用Word2Vec模型训练词向量,以相加平均法合成短文本向量;在此基础上结合基于情感本体所得出的每条短文本的情感值,构建词向量与情感本体相融合的短文本表示模型;最后采用K最近邻分类算法完成短文本情感分类。相比传统的基于词向量、基于情感本体或其他单一技术路线的分类方法,词向量与情感本体相融合的分类方法在准确率、召回率、F1值均有明显的提升。

At p re se n t ,two ways are mainly adopted for short tex t sentiment classification： statistical Natural Language Processing and emotional semantic characterist ics, while the researches on the combination of the two methods is few. T h u s ,this paper will design a classification method th at is basedon the combination of word vector and emotional ontology is designed in this paper. F i r s t ly, the wordvector was trained by Word2Vec model and short text vector was synthesized by adding averaOn this b a s is, short tex t expression model which integrates word vector and emotional ontology was constructed by combining emotional value of each short te x t. E ven tua lly, short tex t sentiment classification was completed by using KNN algorithm. Compared with the traditional classification methods which are based on the word v e c tor ,emotional ontology, or some other single technical classification method combining word vector and emotional ontology gets an obvious improvement on precision, recall rate and F 1 value.