Topic modeling and emotion classification of Chinese social media based on deep learning method-- Taking the microblog public opinion on gold price in the past year as an example

This paper studies a method of Chinese social media topic modeling and sentiment classification based on deep learning, taking the microblog public opinion of gold price in the past year as an example. analyzed 41,942 micro blog entries related to gold price, using the BERTopic model for text topic modeling, and using the BERT-TextCNN-BiLSTM model for sentiment classification. result displayed that the public discussion on gold price is particularly active in certain months, and the discussion content is mainly focused on six major topics, such as the US dollar and gold market, the geopolitical impact in the Middle East, and trading trading strategies and operations. The emotional analyzed shows that the public has different emotional tendencies towards different topics, among which the attitude of analyzed of the gold retail market is more positive, and the attitude of media reports is more negative. On the whole, the method in this paper effectively captures the changes of public sentiment on social media, which is of great significance for analyzed of the social impact of economic activities.