Research On Text Sentiment Analysis Method Based On Deep Learning

Emotional analysis is rapidly evolving along with the rise of online social media such as reviews,blogs,and forums.This is because today in these online social media,we can obtain huge-scale perspective data that can help us discover and mine useful information.The main purpose of people posting on social media platforms is to express their opinions.Therefore,user-generated content in social media contains a large amount of user opinion information.To extract useful knowledge from it naturally requires research on sentiment analysis,which has been called the core issue of social media analysis.Traditional text sentiment analysis methods are prone to problems of insufficient coverage and single application field.Insufficient samples and difficulty in extracting artificial features will directly lead to poor text sentiment analysis.Deep learning has received wide attention from scholars since it was proposed,and it has achieved many excellent results in the field of natural language processing.Therefore,this paper studies the sentiment analysis algorithm of text based on deep learning.This paper mainly proposes two text sentiment analysis models: one is a text sentiment analysis model based on graph convolutional neural network.It is combined with sentiment keyword extraction algorithm to construct the original text information into a text topological graph expression,and effectively reduce the complexity of composition by accurately extracting emotional keyword information in text data and using community detection algorithms.It can not only achieve better results of text sentiment analysis,but also reduce the calculation cost.Meanwhile,not only the information of the original text is not lost,but also the relationship between the global words can be well reflected.And the graph is used as the input of the graph sentiment analysis model of graph convolutional neural network.The essence of text sentiment analysis is to classify the nodes in the topological graph.In the node classification task,the effect of the graph convolutional neural network analysis model is much better than other existing methods.The other is to use a deep belief network model combined with a multilayer attention mechanism to obtain the feature representation of the text through the multilayer attention mechanism.The deep belief network can retain the original features to the greatest extent possible while reducing the dimension of the features.The combination of the two can Maximize the model's ability to extract features,and finally,through comparative experiments,it proves that it has achieved better classification accuracy.