多特征感知情感分析模型

Multi-Feature Perception Sentiment Analysis Model

情感分析旨在对带有情感色彩的主观性文本进行分析和总结,针对现有的情感分析研究存在语料特征提取不足等问题,提出了一种基于BERT的多特征感知情感分析模型。首先,将情感文本输入BERT进行语料编码,然后使用多个卷积核提取编码文本片段的局部特征设计了局部特征层,再对编码文本进行降采样操作构造全局特征层,通过融合局部特征层与全局特征层得到多特征感知网络,从而增强情感语料间的信息关联。实验表明,所设计的多特征感知情感分析模型是有效的。

Sentiment analysis aims to analyze and summarize subjective texts with emotional colors.Aiming at the problems of insufficient corpus feature extraction in existing sentiment analysis research,a BERT-based multifeature perception sentiment analysis model is proposed.First,the emotional text was input into BERT for corpus encoding,and then multiple convolution operations were used to extract the local features of the short text to design a local feature layer,and then the encoded text was down-sampled to construct a global feature layer.By fusing the local feature layer and the global feature,the multi-feature perception network was obtained,thereby enhancing the information association between emotional corpus.Experiments show that the multi-feature perceptual sentiment analysis model designed in this paper is effective.