Research On Aspect Level Sentiment Analysis Based On Deep Learning

Text sentiment analysis is an important branch in the field of natural language processing.It is a research on the views expressed by entities.Text sentiment analysis covers many fields,such as text mining,information extraction,machine learning,information retrieval and so on.In recent years,with the progress of science and technology and the rapid development of the Internet,people's way of shopping has undergone tremendous changes,online shopping has become a new form of today's society.E-commerce platform provides users with an online comment function to publish users' comments on goods or services.With the popularity of the Internet and the increase of e-commerce platform,the data of comment text on the platform is increasing rapidly and exponentially.How to efficiently and quickly mine information from the massive comment text has become a research hotspot at home and abroad.Traditional text sentiment analysis has a good effect in solving the problem of text sentiment polarity judgment,but with the growth of demand,users hope to get the sentiment tendency analysis of specific attributes of the evaluation object.Therefore,fine-grained sentiment analysis based on aspect level is proposed.Aiming at the problems to be solved in aspect level sentiment analysis,this paper proposes a pre training model based on Bert(Bidirectional Encoder Representations from Transformers)combined with iterated dilated convolutional neural networks,based on the aspect level emotion classification model of IDCNN and Bi LSTM,an aspect level emotion classification model based on stacking idea is proposed,and the effectiveness of the proposed model is verified on self constructed data sets.The main work of this paper is as follows.:(1)Aiming at the deficiency of the existing data set for aspect level sentiment analysis task experiment,this paper uses the requests module of Python language to crawl the camera comment data of Jingdong Mall,and constructs a camera online comment data set to verify the aspect level sentiment classification model proposed in this paper.(2)Aiming at the problem that text feature extraction based on traditional machine learning technology can not solve the problem of polysemy,this paper proposes to use the pre training model-Bert.In order to solve the problem of feature information Loss in CNN,an iterative expansion convolution network is proposed for feature extraction.In order to solve the problem of imbalanced classification of sample data,the focal Loss function is proposed to replace the traditional cross entropy function as the Loss function of the model.In order to verify the effectiveness of the model,a variety of benchmark models are used for comparative experiments.The experimental results show that the performance of the proposed bert-idcnn-bilstm model in F1 value is better than other models.(3)In order to overcome the disadvantage of using parallel attention mechanism when extracting text features in Bert model,this paper uses stacking idea to fuse LSTM,CNN and bet-idcnn-bilstm emotion classification methods,and carries out corresponding experimental verification on the self constructed data set,The results show that the performance index of F1 value of the proposed model is better than that of other models.