Research On Sentiment Analysis Of Social Texts Based On Deep Learning

Text sentiment analysis is the process of extracting and identifying the sentiment information contained in a text using computer technology,which has excellent research implications and practical application value and has broad potential for application in various fields.With the development of social networks,people start to express their opinions and share their daily lives on the Internet,analyzing the emotional information contained in these contents can obtain valuable reference data for analyzing the direction of public opinion,adjusting marketing strategies and assessing mental health.In recent years,people’s perception of health has changed,and they begin to pay more attention to and value mental health gradually.Therefore,this paper uses deep learning to analyze the emotion of social text,aiming at mining the emotional information contained in the text,helping people understand the changes of personal emotion and mental health,and then do a good job of self-emotion management and psychological intervention.The main research contents of this paper are as follows:Firstly,to address the problems of text vector representation semantic richness and insufficient feature extraction in sentiment analysis tasks,this paper proposes a sentiment analysis method for classifying the sentiment tendencies of social texts and the emotions embedded in social texts,which uses Ro BERTa model to obtain dynamic word vectors to solve the problem of multiple meanings of words that are prone to occur in Chinese texts and obtain rich semantic representations,and then uses Dc Bi GRU to capture the textual single-sentence features and textual global features in both positive and negative directions,the two parts of the information are fused to obtain the final feature information,and the attention mechanism is used to adjust the allocation of weight resources to capture the important semantic feature information.Secondly,in order to enable the proposed model to maintain a good classification performance despite the sample imbalance,the Focal Loss function is used as the loss function of the model to optimize the model,to reduce the weight contribution of classes with a larger proportion of samples in the dataset improves the classification performance of the model on a few classes,and its effectiveness is demonstrated by experiments.Finally,to address the problems of too many emotion options and not enough objective emotion reports in the mood recording function of existing mental health Apps,this paper develops a sentiment analysis We Chat mini program based on the proposed model.The mini program can analyze users’ mood through the text inputted by user and generate mood curve in a specified period of time to help users understand the mood and mental health status.