Text Sentiment Analysis Based On Deep Learning Algorithm

With the advent of the Web2.0 era and with the emergence of social platforms and e-commerce,the number of Internet users has seen an explosive growth and tens of thousands of textual data generated every day.Regarding government departments,research institutes and Internet service providers,we can apply text sentiment analysis to these data reasonably and dig out the true intentions of the users behind the data,which has important research significance and practical value.Traditional text sentiment analysis methods are mainly based on lexicon and machine learning methods.With the increase of data,these methods show some limitations.Since the in-depth learning was put forward in 2006,a large number of research results show that compared with the traditional methods,many models in deep learning can achieve better performance in multiple fields.This paper attempts to apply the relevant model of deep learning to analyze text sentiment.The main tasks are as follows:For the text sentiment analysis based on the convolutional neural network,through the introduction of attention mechanism,when performing the task of the text sentiment analysis,not only the effective local features in the text can be obtained through the convolutional network,but also the word features of the related results in the text can be obtained.Word features,and a text sentiment analysis model combining convolutional neural network and attention model is proposed.Aiming at the problem of long-term dependence cannot be solved in circulatory neural networks,bidirectional cyclic neural networks based on long-term and short-term memory are used in the text sentiment analysis and the input word vectors are optimized.A bidirectional Neural network text sentiment analysis model based on joint training of words is proposed.For the above two models,through the design and implementation of related experiments,the effectiveness of the proposed model is verified.