融合情感特征的卷积神经网络情感分类模型

Convolutional neural network sentiment classification model fused with sentiment feature

深度学习已广泛用于情感分析中,但现有研究往往忽略词语自身的词性和情感特征,导致情感分析的准确率较低。为了解决这一问题,本文提出了一种融合情感特征的卷积神经网络情感分类模型。该方法将词语、词性、词语情感特征向量化后拼接作为模型的输入层,利用多特征融合解决同词不同词义问题以及表达多情感特征信息;采用分段池化,综合语句结构位置信息,提取多种特征。相较于传统的CNN,实验表明,该方法能明显提高情感分类的准确率。

Deep learning has been widely used in sentiment analysis,however,existing studies often ignore the part of speech and emotion features of words,resulting in low accuracy of sentiment analysis.To solve the problem,we propose a method for convolutional neural network sentiment classification model which fused sentiment features.In this method,words,parts of speech and sentiment features are vectorized and pieced together as the input layer of the model,and multi-feature fusion is used to solve the problem of different meanings of the same word and express multi-sentiment feature information;By using segmented pooling method,syntactic structure location information is integrated to extract multiple features.Compared with the traditional CNN,the experimental results show that this method can significantly improve the accuracy of sentiment analysis.