Design And Implement Of User Comment Sentiment Analysis System Based On Deep Learning

With the rise of online video media,the major video websites purchase the copyright of TV dramas to play on their platforms.The broadcast volume of TV dramas directly affects the revenue of video platforms.How to estimate the market value of TV dramas becomes an important issue.Through the sentiment analysis of the user comment data of the film and television drama,the user's preference and the factors affecting the user's preferences can be clearly grasped,and the purchaser can obtain the market sentiment.Therefore,the user comment sentiment analysis system was designed and implemented.Through this system,websites can query about the overall emotions of the users,user emotions for specified episode,emotional trends and other emotional information.The emotional classification and emotional keyword extraction of the film review are the main research contents.Deep learning model is used to solve text sentiment analysis problems.The following three main aspects of the work are completed.(1)In order to improve the accuracy of sentiment classification,the combination of Attention,RNN and CNN in deep learning is explored.A fusion model ARC and its variant M ARC are proposed,the fusion model can further extract n-gram features on the characteristics of bidirectional timing.The sentiment tri-class evaluation is carried out on the public sentiment dataset twitter,yelp and the Film review corpus constructed.The experimental results verified the effectiveness of the fusion method.(2)The attention mechanism can learn the important distribution of words in sentences.In order to improve the interpretability of emotion classification,an optimized keyword extraction method was proposed by utilizes the attention output in ARC model.On the SemEval dataset,compared with the traditional TextRank,TF-IDF,and keyword extraction algorithm based on complex networks,the results show that the F1-measurement,accuracy and recall have at least 1.1%improvement.(3)the user comment sentiment analysis system based on deep learning was designed and developed.The system mainly implements five modules:data collection,sentiment analysis,keyword extraction,emotion retrieval and visualization.The sentiment classification algorithm and keyword extraction algorithm are applied to the corresponding modules of the system.After functional and performance testing,the system meets user needs.