FinBERT-FOMC-aspects

This model is a specialized version of ZiweiChen/FinBERT-FOMC, further fine-tuned to detect sentiment across specific economic aspects: growth, employment, and inflation. It is designed to handle the nuanced "hallucinated logic" and complex discourse found in Federal Reserve communications and FOMC minutes.By using an Aspect Selection Mode, the model allows researchers to query sentiment relative to a specific economic indicator rather than just receiving a general sentence-level score.

Labels

The model outputs three categorical sentiment labels:

• 0: Negative (e.g., "Growth is slowing unexpectedly")
• 1: Neutral (e.g., "Inflation remained steady")
• 2: Positive (e.g., "Employment figures showed robust gains")

Aspect Selecton Mode

Please choose mode one of 'growth', 'employment', 'inflation'

from transformers import pipeline

mode = "inflation"
sentence = "Inflation has been persistently high over the last few months, requiring tighter monetary policy."

# The pipeline can download and load directly from a subfolder in one line
clf = pipeline(
    "text-classification", 
    model="Wonseong/FinBERT-FOMC-aspects", 
    subfolder=f"finbert-fomc-{mode}-v4"
)

# text-classification pipelines reliably return [{'label': '...', 'score': ...}]
res = clf(sentence)[0] 
labels = {"LABEL_0": "Negative", "LABEL_1": "Neutral", "LABEL_2": "Positive"}

print(f"Aspect   : {mode}")
print(f"Sentence : {sentence}")
print(f"Sentiment: {labels.get(res['label'], res['label'])} (confidence: {res['score']:.4f})")

Related Work

• Kim, W., & Lee, C. L. (2024). Bounded rationality in central bank communication. arXiv preprint arXiv:2411.04286.

• Kim, W., Spörer, J., Lee, C. L., & Handschuh, S. (2024, November). Is small really beautiful for central Bank communication? Evaluating language models for finance: Llama-3-70B, GPT-4, FinBERT-FOMC, FinBERT, and VADER. In Proceedings of the 5th ACM International Conference on AI in Finance (pp. 626-633).

• Gössi, S., Chen, Z., Kim, W., Bermeitinger, B., & Handschuh, S. (2023, November). Finbert-fomc: Fine-tuned finbert model with sentiment focus method for enhancing sentiment analysis of fomc minutes. In Proceedings of the Fourth ACM International Conference on AI in Finance (pp. 357-364).