Research On The Sentiment Analysis Of Takeout Comments Based On Active Learning

In recent years,the application of Internet local life style has developed rapidly,and the takeout service has provided great convenience for the public,especially the young people.At the same time,more people are willing to express their attitudes and views on the Internet,so the number of takeout comments has increased dramatically.Under the massive data,it is unrealistic to rely on manual analysis,and so more intelligent active analysis technology has developed rapidly.This paper focuses on the takeout comment text,and uses active learning technology to mine the emotional orientation and theme of the text,thereby helping businesses and platforms master user needs.This paper mainly has three innovative points.Firstly,in terms of model,improve the original model and establish a Text RCNN(TRCA)model with self attention mechanism.Secondly,build a framework of emotional analysis model that combines active learning and deep learning,using different models in the sample selection stage and prediction stage.Finally,a new measurement index is constructed based on a sample selection method combining uncertainty and sample diversity.In all,a Text RCNN(TRCA)emotion analysis model with self attention mechanism based on fusion of uncertainty and sample diversity is proposed.This paper uses data to verify the advantages of above innovation points.This paper uses takeout review text data,preprocesses it and introduces the basic situation of the text,including word frequency statistics.Secondly,through comparison with multiple models and original strategies,it is finally found that the TRCA model proposed in this paper,which integrates uncertainty and sample diversity in the selection strategy under active learning method,has a high prediction accuracy and indeed has its advantages.After that,based on the LDA theme model,the topic modeling and analysis of the takeout sales comment text are carried out.According to the confusion index,the optimal number of topics is determined to be three,and their corresponding topic keywords are found.According to the keywords,the topics are summarized and analyzed as: delivery,food quality,and feedback service.Under the theme model,combined with the method of emotion analysis,the emotion of comments under different themes is statistically analyzed.It is found that users have more positive comments on the delivery service,and are more satisfied with the service.However,in terms of food quality and feedback service,there are many negative comments,and users tend to have negative emotions.Finally,according to the results of sentiment analysis and theme analysis,some suggestions are put forward for businesses and platforms.