A BERT-Based Method for Mining and Quantifying Online Consumer Public Opinion

With the popularity of Internet and social media, the expression and dissemination of public sentiment on social platforms have become increasingly frequent. In this paper, BERT is applied to sentiment analysis, and a new online Consumer Public Opinion Index (CPOI) is proposed to quantify consumer sentiment fluctuations in social media. The study found that the accuracy and robustness of BERT significantly superior to TextCNN and LSTM in consumer sentiment analysis. Further analysis shows that there is a significant long-term equilibrium relationship between CPOI and traditional Consumer Confidence Index (CCI), which can reflect the changing trend of consumer confidence. This method has strong real-time monitoring ability, and provides a new idea for the future dynamic prediction of consumer confidence. The results provides the technical support for social media public opinion minging and market forecast, and shows the potential of BERT for sentiment analyzing and public opinion quantifying.