Text Sentiment Analysis Based On Residual Network With Text Similarity

In today’s computer network age,application platforms such as Weibo,Taobao,and JD has brought great convenience to people’s lives.At the same time,a large number of comments and opinions have emerged from this.The user’s comments and opinions represent the user’s emotional inclinations.By mastering the user’s emotional inclinations,you can effectively recommend or avoid certain topics or products for users.How to dig out important emotional information from these contents is particularly important.Aiming at the task of sentiment analysis of text,this paper proposes a residual network model of sentiment analysis,which makes full use of the feature relevance between similar texts and has achieved good results on sentiment analysis tasks.The work of this paper mainly includes the following points:1.Based on deep learning theory and text similarity ideas,a residual network text sentiment analysis model based on the introduction of text similarity is proposed,which realizes the sentiment tendency analysis of texts.2.Based on the transfer learning method,the SBERT(Sentence Bidirectional Encoder Representations from Transformers)model is used to construct equal-length text vectors that contain the semantic information of the text and facilitate the calculation of text similarity,and the BERT(Bidirectional Encoder Representations from Transformers)model is used to transform the text data to the semantic matrix.3.Based on the principle of text similarity,this paper proposes a method of combining the weighted semantic matrix of the text to be predicted and its similar text to represent the text to be predicted,which enhances the semantic and emotional relevance between similar texts.4.By constructing the residual network model,fine-tuning the network structure,and adding a self-attention mechanism,the network model can better learn the emotional characteristics of text and the correlation characteristics between similar texts.This paper conducts sentiment analysis experiments based on the Tan Songbo hotel review data set and IMDB data set.The experiment shows that the text sentiment analysis model proposed in this paper has achieved better results on these two data sets,and the accuracy rate is higher than some existing ones.Sentiment analysis algorithms and models.