- Project
- AI/ML
- Solo
- 2025
Financial news sentiment: RoBERTa vs FinBERT
A general-purpose model beat the domain-specific one on 3-class financial sentiment.
How it works
- 01877 labelled test headlines, 3 classes
- 02BiLSTM, RoBERTa, FinBERT fine-tuned on the same split
RoBERTa accuracyWeighted F1 0.835 on 877 test headlines; FinBERT 81.0%, BiLSTM 59.1%.
RoBERTa accuracy
Weighted F1 0.835 on 877 test headlines; FinBERT 81.0%, BiLSTM 59.1%.
The problem
Does a finance-specific language model beat a general one on financial headlines?
My part
Fine-tuned and compared all three models.
What did not work
- BiLSTM reached only 59.1%.
Limitations
- One dataset and one split; no significance test.