Technology

Model

Using a fine-tuned FinBERT model specialized for Federal Reserve communications, we measure the sentiment expressed in FOMC statements. The model identifies subtle changes in policymakers’ tone toward the economy, producing a continuous index that reflects evolving confidence or concern.

Data & Transforms

• FOMC sentiment: economic tone extracted from FOMC statements, displayed as a smoothed measure of recent sentiment.
• Forward returns: Fwd12m = Close[t+12] / Close[t] − 1. Recent months can be null until future prices exist.

App

• Backend: Flask, pandas; /data serves compact JSON.
• Frontend: ECharts, Tailwind + DaisyUI.

Notes

• Sentiment ≠ signal; markets respond to many factors beyond policy language.
• Last ≈12 rows of forward returns can be null until future prices exist.
• Caching is in-process with a configurable TTL.