Design Of Emotion Analysis System For Social Platform Based On Deep Learning

With the rapid development of Internet technology,social platforms such as microblogs are becoming more and more mature.The rapid increase of users leads to the explosive growth of text data.The comments made by users are short and heavily colloquial.This makes it more difficult to accurately identify the emotional information hidden in the text.Traditional sentiment analysis methods mainly rely on the manual construction of emotion dictionary.It is time-consuming and laborious to select different feature selection methods for corpus in different fields.It has been unable to meet the needs of short text sentiment analysis.Therefore,this paper combines the method of deep learning.An emotion analysis system of social platform based on deep learning is developed.The main research work is as follows:1)Data is acquired and labeled the data with categories.The list of stop words adopts“ Harbin Institute of Technology Stop Words Vocabulary”.One part of the dataset is downloaded directly from website and analyzed.We use the Scrapy framework combined with beautiful soup to parse the web page and crawl the short text of microblog comments.Finally,we use the method of manual annotation to complete the collation of another part of the experimental data set.2)An affective analysis model based on deep learning is established.Firstly,the embedded module is used to generate the embedding matrix from the processed experimental data.Then the LSTM is used to capture the long-distance dependence.TextCNN is used to capture dependencies between words.Finally,the specific emotion category of the text is output.On the same data set,it is compared with the emotion analysis model based on TextCNN,the emotion analysis model based on RNN and the emotion analysis model based on LSTM.The results show that the sentiment analysis model based on LSTM + TextCNN has the best classification effect.Its accuracy is91.22%.3)Social platform sentiment analysis system is developed.Microblog short text sentiment analysis is used as an application scenario.The experimental methods in this paper are used as the basis.A text sentiment analysis system is designed and implemented.We design the overall architecture of the system according to the demand analysis.It mainly includes five parts: data acquisition and storage,data processing,model training,classification and prediction,and human-computer interaction.The system is built by Vue,Django,Scrapy framework.It is written in Java and Python.Finally,the system visualization is realized by Echards.