[Purpose/significance] Weibo is a Chinese social media platform based on user relationships. Users can comment on hot topics to express their emotions and emotional tendencies. Mining users' emotion and attitude from Weibo topic comments are helpful to the analysis of Weibo public opinion analysis, which has practical significance.
[Method/process] This paper proposes a sentiment analysis method that combines the bidirectional gated recurrent unit neural network (BiGRU) with the attention mechanism to construct an emotion classification model that takes into account training speed and accuracy. Experiments on public datasets show that compared with the classical neural network model, the model in this paper can effectively improve the accuracy of sentiment classification. Experiments are carried out on topics comment data of 10 different fields, and the results verify the feasibility and generalization ability of the model.
[Result/conclusion] Finally, a quick and effective sentiment prediction is made for the comments on hot topics, and the opinions and attitudes of netizens on hot topics are analyzed according to the calculated sentiment index and word cloud graph, so as to grasp the public’s emotional attitude towards public opinion events in time and provide a reference for the subsequent application of popular public event prediction and public opinion analysis and research.
[Limitations] There are also some shortcomings in this study, such as errors in sentiment analysis of comment sentences with complex semantics such as irony, and no fine-grained sentiment analysis, which still needs further research and improvement.