A FinBERT Framework for Sentiment Analysis of Chinese Financial News
Yuesheng Huang, Renyi Lin, Jiawen Li, Yiqi Li, Jiajia Chen, Jiaqi Liang, Guanyuan Feng, Jujian Lv · 2024
In the era of big data, sentiment analysis of the Chinese stock market has become a research hotspot. As a vital reflection of market sentiment, financial news has a significant impact on stock market trends. In this paper, we propose a framework based on the FinBERT pre-trained model optimized for financial news sentiment analysis in China. In particular, the model utilizes the Bidirectional Encoder Representations from Transformers (BERT) to deal with semantics, syntactic features, and specialized vocabulary of financial texts through self-supervised learning. The fine-tuning stage further optimizes the model, so that it can accurately predict sentiment tendencies. Experimental results show that the proposed model outperforms others on the sentiment classification task, with an accuracy of up to 94.52%. Consequently, the model can provide participants in the financial industry with a valuable reference of market trends, which helps investors and analysts make decisions properly.