The School's Center for Business and Social Analytics (CBSA) has developed a fintech application project titled FinSent. It is a free-to-use financial sentiment analysis web portal that automatically tracks sentiment expressed in financial reports and earnings calls of companies listed in the United States and Hong Kong.

Leveraging a deep-learning-based natural language processing (NLP) model, the portal provides investors with a new perspective to analyze listed companies beyond financial information. Leaders of the project team, Prof. Allen HUANG, Associate Dean (UG Programs), and Prof. YANG Yi of the Department of Information Systems, Business Statistics and Operations Management, spoke to local media about how the project unlocks the power of sentiment analysis in understanding corporate disclosures.

FinSent is powered by FinBERT, a BERT model (Google’s machine learning framework for NLP) developed by CBSA and trained on financial communication corpus of US-listed firms, with total size of 4.9B tokens. FinBERT outperforms the BERT model in accuracy when classifying sentiments in financial text.

Given investors’ strong interest in business outlook, the project team is looking to roll out a sentiment score for forward-looking statements in the near future. Prof. Huang said that, going forward, as the team better grasps market needs, it may explore the potential for commercialization by providing application programming interface, and customized data analytics services to users.

Media reports (Chinese only):
Ming Pao
Sing Tao Daily

Access the FinSent portal here.