Prosus AI / Dogu Araci · Finance
FinBERT
A BERT model fine-tuned on financial communication text for accurate financial sentiment analysis and opinion mining.
Overview
FinBERT is a pre-trained language model built on top of BERT and further trained on a large corpus of financial text including corporate reports, analyst notes, and financial news. It is the most widely adopted model for financial sentiment analysis, capable of classifying text as positive, negative, or neutral with high accuracy. FinBERT has become the standard benchmark model for financial NLP research.
Parameters
110M
Architecture
BERT-Base
Training Data
Financial PhraseBank + TRC2 financial corpus
Context Window
512 tokens
License
Apache 2.0
Capabilities
Financial sentiment classification
Opinion mining from financial text
Tone analysis of corporate communications
Market sentiment scoring
Use Cases
Analyzing sentiment in earnings call transcripts
Scoring market sentiment from financial news feeds
Monitoring social media for investment-relevant opinions
Detecting tone shifts in corporate disclosures
Pros
- +Industry standard for financial sentiment analysis
- +Lightweight and easy to deploy on minimal hardware
- +Well-documented with extensive academic validation
- +Open-source and free for commercial use
Cons
- -Limited to sentiment analysis; not a general-purpose model
- -Small context window restricts analysis of long documents
- -Encoder-only architecture cannot generate text
- -Pre-trained on English financial text only
Pricing
Free and open-source. Available on Hugging Face. Can run on CPU for inference making it extremely cost-effective.