Research And Implementation Of Emotion Recognition System For Micro-blog Short Texts

As the rapid development of the Internet, people more and more frequently communicate through the Internet. Social platforms such as Sina Weibo, RenRen, Twitter and Facebook further enhance the spread of information between people. People can communicate with others through the Internet anytime and anywhere, participate in various topics of discussion and express their own comments and opinions. These text data contain great value in both research and industry. Text emotion analysis is a popular research field nowdays. In this paper, we mainly study the emotion tendency anlaysis in emotion recognition, which is carried out by using statistics, machine learning and deep learning methods to discriminate the emotional tendencies of the text and obtain the emotional expression of the users.This paper mainly studies the word vector model and deep learning model on text classification filed, including Convolution Neural Network(CNN) and Long Short-Term Memory (LSTM): We proposed an new word vector model called Order_w2v considered word order feature based on Word2vec model and an end-to-end CNN-LSTM model for text classification based on the advantages of CNN and LSTM, aiming at improving the classification accuracy. Through the real data experiments,it is shown that deep learning model such as CNN, LSTM and CNN-LSTM can achieve a higher classification accuracy taking Order_w2v as the model feature input than word2vec. In the meantime,CNN-LSTM model can achieve higher classification accuracy in text classification than the other mainstream deep learning models including CNN and LSTM.This paper designs and implements the emotion orientation identification system for microblogging short text based on Order_w2v and CNN-LSTM model. The contents contain the following aspects:Studying the text style of Chinese micro-blog, implementing micro-blog preprocessing technology;Studying and implementing the new word discovery technique based on micro-blogging through Natural Language Processing (NLP); Implementing the word vector representation on micro-blog and proposing a improved word vector model based word order called Order_w2v; Implementing deep learning models: CNN,LSTM and CNN-LSTM; Implementing the Micro-blog Emotional Tendencies Recognition System based on these models. We complete the Micro-blog Emotional Recognition System test based on micro-blog hot topic data. And the result shows that the system in micro-blog emotional tendencies recognition has a good result.