CrudeBERT: Applying Economic Theory towards fine-tuning Transformer-based Sentiment Analysis Models to the Crude Oil Market
This work addresses the need for more accurate sentiment analysis tailored to the crude oil market, which is incremental as it builds on existing transformer methods with domain-specific adaptations.
The paper tackled the problem of asset-specific sentiment analysis in financial markets by developing CrudeBERT, a model that fine-tunes FinBERT using economic events to improve sentiment classification for crude oil headlines, resulting in outperformance over existing solutions in this domain.
Predicting market movements based on the sentiment of news media has a long tradition in data analysis. With advances in natural language processing, transformer architectures have emerged that enable contextually aware sentiment classification. Nevertheless, current methods built for the general financial market such as FinBERT cannot distinguish asset-specific value-driving factors. This paper addresses this shortcoming by presenting a method that identifies and classifies events that impact supply and demand in the crude oil markets within a large corpus of relevant news headlines. We then introduce CrudeBERT, a new sentiment analysis model that draws upon these events to contextualize and fine-tune FinBERT, thereby yielding improved sentiment classifications for headlines related to the crude oil futures market. An extensive evaluation demonstrates that CrudeBERT outperforms proprietary and open-source solutions in the domain of crude oil.