BiLSTM-CNN Sentiment Analysis Of Weibo With Expression Fusion

As a very popular online social platform,Weibo covers a wide range of discussion topics,from current affairs news,medical care,entertainment to academic aspects.The people participating in the topic also include all age groups.Netizens can comment on the topics they care about..This has also led to the fact that Weibo contains a large amount of text information.Through sentiment analysis on the comment text of Weibo,on the product side,the product side can understand the people's attitude towards the product or planned activities and improve it;in the government On the one hand,the government can understand the people's attitudes towards cyber incidents and the fermentation situation,so as to better guide public opinion.Therefore,text sentiment analysis not only has commercial value,but also plays a positive role in maintaining social stability.The research of text sentiment analysis is a very meaningful subject.The main research work of this paper is as follows:First,a method of vectorization of emoticons is proposed.Firstly,a data set of emoticons was constructed,and positive and negative samples of descriptive sentences were constructed.Secondly,the word vector tool was used to model the probabilistic similarity of emoticon names and descriptive sentences to obtain emoticon vectors.Second,a serial model of Bi LSTM-CNN is constructed.Due to the polysemous situation in Chinese text,a bidirectional Bi LSTM model is introduced to extract text features that contain context information.The structure of Chinese text is relatively complex.The CNN model can extract important features,and the CNN model is introduced to extract local semantic features of the text.Other studies serialized the two models.First,CNN was used to extract the local semantic features,and then the Bi LSTM model was used to extract the context information of the local features,which would still cause the lack of information.In this paper,the two models are paralleled and both features are extracted at the same time.,To make the characteristics of the input model more complete.Thirdly,the validity of the model is verified by using the data set comparison.In terms of emoticons,we first use the basic CNN method to verify it,and then supplement the comparison verification on the Bi LSTM-CNN serial model.In the Bi LSTM-CNN serial model,the basic Bi LSTM model and the CNN model are used for comparison and verification.