Research On Aspect Level Emotion Analysis Technology Based On Deep Learning

With the development of the Internet,people are more and more used to expressing their opinions and emotions on the internetwork.And Collecting,analyzing and sorting out these text information with emotional tendency can better understand people’s behaviors and habits which have certain social and commercial value.Traditional sentiment analysis focuses on documents or sentences.Traditional sentiment analysis mainly focuses on documents or sentences,while aspect level sentiment analysis can focus on a certain aspect of the text for fine-grained analysis,which has become a research hotspot in the field of sentiment analysis in recent years.The main method of emotion analysis is to apply deep learning method,fully excavate semantic features in text and establish the relationship between words.For aspect level affective analysis task,the main research work and innovation of this paper are as follows:1、Through aspect level sentiment analysis,it is found that the previous method of using LSTM network to model aspect words and context can not distinguish the importance of words in the text,and lack of attention to aspect words and important words in the text.To make up for these shortcomings,this paper proposes an aspect level sentiment analysis model combining attention mechanism and LSTM network.The model uses multi-head self-attention mechanism and aspect specific attention mechanism to process the hidden layer data of LSTM network output;And the interactive attention mechanism is added to obtain the semantic connection between text and aspect words.Then the semeval2014 evaluation task is used to evaluate the model,and the comparative experiments are carried out on the restaurant and lap datasets respectively.Experimental results show that the proposed model achieves good classification results in aspect level sentiment analysis task,and the accuracy and F1 value are significantly improved.2、Based on the model and method proposed in this paper,a text sentiment analysis system for restaurant reviews is developed.The system can extract the attributes and emotional polarity of the text input by users,display the results of extraction and polarity prediction,and process multiple texts,and display the results visually.