Is Domain Adaptation Worth Your Investment? Comparing BERT and FinBERT on Financial Tasks

Bo Peng, Emmanuele Chersoni, Yu-Yin Hsu, Chu-Ren Huang


Abstract
With the recent rise in popularity of Transformer models in Natural Language Processing, research efforts have been dedicated to the development of domain-adapted versions of BERT-like architectures. In this study, we focus on FinBERT, a Transformer model trained on text from the financial domain. By comparing its performances with the original BERT on a wide variety of financial text processing tasks, we found continual pretraining from the original model to be the more beneficial option. Domain-specific pretraining from scratch, conversely, seems to be less effective.
Anthology ID:
2021.econlp-1.5
Volume:
Proceedings of the Third Workshop on Economics and Natural Language Processing
Month:
November
Year:
2021
Address:
Punta Cana, Dominican Republic
Editors:
Udo Hahn, Veronique Hoste, Amanda Stent
Venue:
ECONLP
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
37–44
Language:
URL:
https://aclanthology.org/2021.econlp-1.5
DOI:
10.18653/v1/2021.econlp-1.5
Bibkey:
Cite (ACL):
Bo Peng, Emmanuele Chersoni, Yu-Yin Hsu, and Chu-Ren Huang. 2021. Is Domain Adaptation Worth Your Investment? Comparing BERT and FinBERT on Financial Tasks. In Proceedings of the Third Workshop on Economics and Natural Language Processing, pages 37–44, Punta Cana, Dominican Republic. Association for Computational Linguistics.
Cite (Informal):
Is Domain Adaptation Worth Your Investment? Comparing BERT and FinBERT on Financial Tasks (Peng et al., ECONLP 2021)
Copy Citation:
PDF:
https://aclanthology.org/2021.econlp-1.5.pdf
Video:
 https://aclanthology.org/2021.econlp-1.5.mp4