Research On Sentiment Analysis Model Based On Weibo Topic

Chinese Weibo sentiment analysis aims to extract the opinions and tendencies of netizens towards hot current events.Through the emotional analysis of social platform text,we can get the emotional state of netizens,which plays an important role in many aspects,such as government public opinion control,company product marketing,public opinion survey,etc.Nowadays,with the development of Internet technology,netizens rely more and more on social platforms such as Weibo and WeChat.The practical significance of emotional analysis of Chinese Weibo has become more and more important.But in the current network environment,the words which are used by people to express their views and emotions become more and more,and the ways become more and more abundant.A sentence maybe contains Chinese,English,network catchwords and emoticon concurrently.In addition,the Internet buzzwords tend to be ironic.The traditional sentiment analysis model has gradually been unable to meet the needs of sentiment analysis in the current Weibo environment.This thesis will improve the traditional sentiment analysis model based on the current Weibo environment,and propose a fine-grained sentiment analysis model based on Weibo topic.The current work in this thesis mainly includes three parts:1.Based on the traditional Chinese Weibo sentiment analysis model,combined with the idea of aspect-level sentiment analysis,we have improved the original model and proposed a new model to make it competent for the emotional analysis task based on Weibo topic words.2.Based on the new model proposed in this thesis,we improve the model by improving the attention mechanism strategy and adding linguistic knowledge to improve the accuracy of the model.3.Taking the new model as the core and combining with the practical application environment,the system design and implementation of the model are completed by using front-end and back-end programming technology.For the first and second part of the research theme,corresponding comparative experiments are designed.The experimental results show that the model proposed in this paper has good performance on given data sets and high performance in topic-based Weibo sentiment analysis tasks.