A Study on the Sentiment Analysis of Netizen's Response to Subjects.

This study proposes a sentiment analysis system to explore the response and the preference of netizens on special social network subjects. Our proposed system consists of 4 procedures: (1) crawls a large amount of network articles and messages for data collection; (2) proposes a vocabulary encoding model based on semantic text encoding method (word embedding) for text processing; (3) develops an emotional learning classification model based on word2vec and support vector machine to classify articles into three emotions; (4) implements sentiment analysis based on emotional learning classification model to obtain the preference of the subjects. Our system can effectively classify the messages of different subjects into different emotional categories to observe the netizensu0027 response to the subject, and further understand the subjectu0027s attention and preference. Moreover, our system also focuses on the comparison of preferences between multiple subjects to understand whether a subject is more popular with netizens than other subjects. Then, we can clearly observe and grasp the influence of these subjects.