Research And Implementation Of Sentiment Analysis System For Microblog Reviews

In an age of information explosion,the internet has become increasingly relied upon and as technology advances social media platforms are becoming more and more popular.Emerging social media platforms such as Weibo and We Chat are popular with users who can easily and quickly post text,pictures and videos to facilitate the exchange of information between them.People often comment on microblogs to express their personal opinions and to get instant access to information.Analysing microblog comment data helps the government to quickly understand the current public opinion environment.This allows the government to provide timely and positive guidance on public opinion,crisis communication,etc.Most of the traditional sentiment analysis algorithms are based on word attention mechanism combined with Bi LSTM algorithm,which results in such methods relying too much on the accuracy of word separation and not focusing more on the sequence relationship within the text.Traditional sentiment analysis algorithms are more likely to target medium-length texts with large paragraphs and standardised wording.The traditional sentiment analysis algorithms are not suitable for microblog comments because they are short and concise,and there are a large number of emoticons and spam messages.This paper proposes to classify comments into three types: spam comments,subjective comments and objective comments.By using feature extraction and machine learning methods to strip and extract spam,subjective and objective comments,the two main categories of spam and objective comments are eliminated to reduce the noise in the text.The subjective comments were then cut into word vectors and a sentiment analysis algorithm was constructed using Bi LSTM combined with a self-attentive mechanism to predict the sentiment polarity in microblog comments by obtaining sentiment categories with a softmax classifier.The final results were obtained by experimenting with the public dataset provided by the Chinese Information Society,and the word vector combined with Self-Attention and Bi LSTM algorithms were more suitable for sentiment analysis of microblog comments.The accuracy of the algorithm was improved by 1.65% compared to the traditional sentiment analysis algorithm.This paper focuses on the sentiment analysis of microblog comments and the alarming operation of microblogs with preset warning values to help government officials and early warning system administrators to better understand the trend of public opinion and respond quickly to public opinion information.After a user satisfaction survey,most users are satisfied with the effectiveness of the system.Therefore,the word vector combined with Self-Attention and Bi LSTM microblog public opinion analysis and early warning system proposed in this paper has high reliability.