Skip to main content

FinBERT Sentiment

In short

AI model trained on financial text — classifies news as positive, negative, or neutral

FinBERT is an AI model that reads financial news and rates it as positive, negative, or neutral. Unlike a general AI, it was specifically trained on earnings calls, analyst reports, and financial news — so it understands financial language.

FinBERT is a fine-tuned version of BERT trained on financial text. It outperforms general-purpose models on financial sentiment tasks. Outputs probability scores for positive, negative, and neutral classifications, which can be aggregated across multiple news articles.

Related concepts

  • News Sentiment — News sentiment combines the emotional tone of recent news articles about a company. Lots of positive headlines = positive sentiment. A string of bad news = negative. It can signal upcoming price moves before they happen.
  • Sentiment Score — The sentiment score blends news and social media signals into one number. Think of it as a thermometer for how the market 'feels' about a stock right now. Positive = optimistic; negative = pessimistic.
  • Cross-Source Agreement — When both news AND social media are positive (or both negative), that agreement makes the signal much stronger than if they disagree. Like getting the same weather forecast from two independent meteorologists.