Emotion Recognition And Classification Research Towards Micro-blog

With the rapid development of the Internet,social network has become an important carrier for people to disseminate information in their daily life.The micro-blog platform is a typical representative of the social network applications.The sentiment analysis for microblog text aims at identifying the emotional categories that micro-blog contains,such as pleasure,sadness and surprise,through analyzing and mining the micro-blog texts published by users.Emotional analysis is a basic research task in the field of Natural Language Processing,which has been closely watched by scholars in this field in recent years.This paper focuses on the research of micro-blog text emotion recognition and classification methods,including the following three aspects:First of all,this paper proposes a method of micro-blog emotion recognition based on the character of word fusion.This method can fully consider the characteristics of microblog language,and fully utilize the combination of characters and words to improve the performance of emotion recognition.In order to distinguish between words and characters of micro blog,this article uses the Long Short-Term Memory Network(LSTM)respectively the words from the text features representation and word study of hidden layer features said.Finally,through the fusion of the two groups of hidden layers,the fusion characteristics of words and characters were obtained and emotional identification was carried out.The experimental results show that this method can effectively integrate characters and words,which is better than traditional emotion recognition method.Secondly,this paper proposes an unbalanced emotion classification method based on multi-channel LSTM.This method aims to solve the problem of unbalanced sample number of emotional categories in emotional classification tasks.The imbalance of the sample distribution will result in the application of traditional machine learning classification methods,which can greatly reduce the classification performance.Firstly,the method proposed in this paper uses the under-sampling technology to obtain multiple groups of balanced training corpus.Secondly,a LSTM model was studied with each training corpus.Finally,the final classification results are obtained by fusing multiple LSTM models.The experimental results show that this method can not only make full use of the training sample,can also maintain the balance between all kinds of different emotions,in the sentiment classification task performance is superior to the traditional unbalanced classification method.Finally,this paper proposes an emotional classification method based on bilingual information.This method is aimed to solve the problem of the lacking of sample and shortness of text and shortness of information in emotional classification task.In particular,the method firstly uses the machine translation program to translate the source language materials into the translated corpus.Secondly,the corpus of corresponding language is merged and expanded,and the corpus of two groups of different languages is obtained.Finally,the text uses the source language and the translation language respectively to carry out the feature representation,and the two-channel LSTM model is established to combine the two sets of features to obtain the fusion characteristics and build the emotional classifier.The experimental results show that the method has a significant improvement in the performance of emotional classification on the micro-blog and Twitter corpus.