Construction And Application Of WBERT-BILSTM Emotion Analysis Model

"Accelerating digital development and building a digital China" is a national strategy proposed by my country in the "14th Five-Year Plan",which points out the direction for the development of my country's digital economy.With the continuous development and improvement of new infrastructure construction,new innovative breakthroughs have been made in digital technology,and China's Internet has shifted from a high-speed development stage to a high-quality development stage.The scale of netizens has grown steadily,rural and elderly groups have accelerated their integration into the online society,the per capita online time has continued to grow,the use of Internet terminal equipment has become more diversified,and social media has gradually become more abundant.More and more user groups share their living conditions and daily habits on a variety of social platforms with the help of a large number of different types of Internet terminals.Weibo can be said to be the most frequently used social media platform by netizens at present.After practical investigation and analysis,we found that Weibo updates a large amount of data information every day.The emotional characteristics involved can be mastered,so as to provide good help for our country to deal with various public emergencies and to deal with public emotions efficiently.This article mainly involves the creation and practical application of microblog text sentiment analysis model.First of all,the research objects involved in this study are mainly the various text data of the users involved in the Weibo platform.With the help of deep learning methods and language models,the data can be integrated and analyzed,so that it can be more Efficiently classify emotions.When implementing sentiment analysis,the most commonly used methods are: learning on the basis of dictionaries and machine learning.The selected dictionaries usually have a direct impact on the effect of dictionary classification methods.The previous machine learning model is obviously unable to meet the needs of the classification work in the current practical operation,and there is a clear gap between the deep learning method and the understanding of the connotation of the entire article.Therefore,in the sentiment analysis task of this article,the in-depth analysis is carried out on the premise of the BERT(Bidirectional Encoder Representation from Transformers)deep learning language model.First of all,fully combine the actual situation and needs of various aspects to select the data,keep the key data involved in the article,and create a classification model that can accurately and comprehensively judge the sentiment of the data based on the situation of the data.Secondly,this article mainly uses the Weibo platform to collect information and data,and provide the necessary training information for model creation and model improvement.After completing the creation of the WBERT model,the results obtained after the model runs and calculates are used as the basis for subsequent experiments.When performing sentiment classification,in order to fundamentally guarantee the accuracy of the information data provided by the WBERT model,this article adds a special bidirectional long and short-term memory network and an attention-dependent sentiment classification module under the premise of the WBERT model.to complete the weighting of sentiment results.Then,the results are applied to the activation function to guarantee sentiment classification,and finally a complete WBERT-Bi LSTM model can be created.In order to analyze the comprehensiveness of the model,a number of comparative experiments were used.Second,this thesis applies models to specific events.First of all,in terms of social public opinion application,we crawl the relevant comments on the incident of "Chongqing bus falling into the river" through the Weibo API,and preprocess and manually mark the corresponding Weibo posts and user comments.corpus for model evaluation.Secondly,in terms of product review applications,the Chinese datasets of laptop and restaurant reviews in Sem Eval2014 and 2016 are manually annotated and then input into the model.By elaborating the events and analyzing the results,it is found that the model can better realize the sentiment classification of social public opinion events and commodity reviews.