Negative Emotions Of Online Users' Analysis Based On Directional Long Short Term Memory

In the era of Internet,network public opinion events are very easy to occur.Although the causes of various network public opinion events are different,the peak of public opinion development is generally accompanied by the generation of negative emotions of netizens.The negative emotions such as disgust and anger pose a great threat to social stability.Therefore,the effective automatic identification of negative emotions of netizens has attracted great attention from various government agencies,and is also a research hotspot in academia.This paper constructs a model of online users' negative emotion recognition,which uses word embedding model to learn word vectors with semantic and grammatical information as text representations,and extracts sentence features from a bidirectional long short term memory(BiLSTM).Joint attention model makes the model pay more attention to the important features of sentences,and realizes the recognition of five emotions of netizens: non-negative,disgust,anger,sadness and fear.An empirical study is conducted with the data of Chinese and English mico-blog data.Experiments show that the word vectors trained by Skip-Gram model are more suitable for sentiment analysis tasks of this paper;BiLSTM learns word features better than LSTM;the attention model can improve the effect of the original model in learning text features;and the online users' negative emotion recognition model can accomplish the netizens' negative emotion multi-classification task well and has better performance than other neural network models or machine learning models.On the basis of judging the polarity of netizens' emotions,the classification strategy of further recognizing netizens' multiple negative emotions is better than that of directly judging netizens' multiple negative emotions.The innovation of this paper lies in the establishment of a netizen's negative emotion recognition model based on negative emotion perspective and deep learning method,and the validity of the model is verified in a variety of language corpus.From the perspective of netizens' negative emotions,this paper studies the identification of netizens' multiple emotions in network public opinion events.The technological innovation of automatic emotional recognition technology is achieved in the field of emotion analysis of online users by applying BiLSTM combined with attention mechanism.