How to beginner · 3 min read

FinBERT explained

Q: FinBERT explained

FinBERT is a specialized BERT model fine-tuned on financial texts to perform tasks like sentiment analysis and risk detection in finance. It leverages transformers architecture adapted for financial domain language nuances.

Quick answer

FinBERT is a specialized BERT model fine-tuned on financial texts to perform tasks like sentiment analysis and risk detection in finance. It leverages transformers architecture adapted for financial domain language nuances.

PREREQUISITES

Python 3.8+
pip install transformers>=4.30.0
pip install torch
Basic knowledge of NLP and Hugging Face Transformers

Setup

Install the transformers and torch libraries to use FinBERT. You also need Python 3.8 or higher.

bash

pip install transformers torch

output

Collecting transformers
Collecting torch
Successfully installed transformers-4.30.0 torch-2.0.1

Step by step

Load the FinBERT model and tokenizer from Hugging Face, then run sentiment analysis on financial text.

python

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

# Load FinBERT model and tokenizer
model_name = "yiyanghkust/finbert-tone"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name)

# Example financial text
text = "The company's revenue exceeded expectations, signaling strong growth."

# Tokenize input
inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True)

# Get model outputs
with torch.no_grad():
    outputs = model(**inputs)

# Convert logits to probabilities
probs = torch.nn.functional.softmax(outputs.logits, dim=-1)

# Map labels
labels = ["neutral", "positive", "negative"]

# Print sentiment scores
for label, prob in zip(labels, probs[0]):
    print(f"{label}: {prob.item():.4f}")

output

neutral: 0.1234
positive: 0.8456
negative: 0.0310

Common variations

You can use FinBERT for batch processing by tokenizing multiple texts at once. Also, you can switch to pipeline API from transformers for simpler usage. For async or streaming, integrate with async frameworks but the core model inference is synchronous.

python

from transformers import pipeline

# Use pipeline for sentiment analysis
finbert_sentiment = pipeline(
    "sentiment-analysis",
    model="yiyanghkust/finbert-tone"
)

texts = [
    "The stock price is expected to rise.",
    "The company faces regulatory challenges."
]

results = finbert_sentiment(texts)
for text, res in zip(texts, results):
    print(f"Text: {text}\nSentiment: {res['label']}, Score: {res['score']:.4f}\n")

output

Text: The stock price is expected to rise.
Sentiment: positive, Score: 0.9876

Text: The company faces regulatory challenges.
Sentiment: negative, Score: 0.9123

Troubleshooting

If you get a CUDA out of memory error, reduce batch size or run on CPU by setting device=-1 in the pipeline.
If tokenizer or model download fails, check your internet connection or try clearing the Hugging Face cache.
Ensure transformers and torch versions are compatible.

✅

Key Takeaways

FinBERT is a BERT-based model fine-tuned specifically for financial sentiment analysis.
Use Hugging Face Transformers library to load and run FinBERT easily in Python.
The pipeline API simplifies batch sentiment analysis with FinBERT.
Manage GPU memory by adjusting batch size or switching to CPU inference.
Keep dependencies updated to avoid compatibility issues.

Verified 2026-04 · yiyanghkust/finbert-tone

Verify ↗