Sentiment Intensity Research And Realization Based On Comments Data

With the development of web2.0,the network infiltrates into all aspects of people's life.Comment Website as a very good entry point for the Internet to penetrate into the line,has become a standard of local vertical life.Sentiment Analysis of the user original comment data reflects the great academic value and commercial value.The simple classification of sentiment can not satisfy the increasing demand for classification strength and precision.This paper mainly studies how to quantify the expression of text sentiment,and studies the sentiment intensity closer to the real situation of the user than the simple sentiment classification research,which has more important academic value.This thesis takes the review data of the DianPing comments as the research object,and realizes a text sentiment intensity analysis system through the research.It analyzes the text sentiment tendency from the quantitative perspective.For the input comment,the system analyzes and obtains the sentiment intensity value of the comment and displays it.First of all,in terms of text representation,we presents a method of text representation using a sentiment dictionary.After the research and annotation steps,a sentiment dictionary with eight dimensions of Surprise,Sorrow,Love,Joy,Hate,Expect,Anxiety,Anger has been realized.The floating point of each dimension represents the weight of the dimension.The experimental results show that a text vector based on the sentiment dictionary represents the maximum possible sentiment information of the text.Second,in terms of sentiment learning,we combine the advanced learning technology of the current frontier.Based on the previous work,we proposes a Bi-LSTM+Attention model,which preserves the contextual and sentiment information better while retaining the advantages of other deep learning models.The stability and effectiveness of the L-RCNN model are verified by the experimental results.Finally,in quantitative expression of sentiment intensity,we establishes a new sentiment intensity measurement system,which divides sentiment intensity into ten dimensions,makes a clear distinction between different degrees of sentiment tendencies.