The Architecture

Jazz Analytics is built on a custom data pipeline designed to analyze harmonic structures at scale. Unlike traditional wikis manually edited by users, our insights are generated algorithmically to ensure consistency across thousands of jazz standards.

1. Chord Processing & Data Cleaning

Raw chord data is ingested from open-source jazz standard datasets. We use a custom Python parser to standardize notation, handling complex variations (e.g., converting C7alt or C7#9#5 into standardized pitch classes).

2. Harmonic Analysis Engine

The core of this project is built upon Music21, a toolkit for computer-aided musicology.

  • Key Detection: We analyze the weighted pitch distribution of each song to determine the tonal center.
  • Pattern Recognition: A sliding window algorithm scans the chord progression to identify functional movements, specifically ii-V-I sequences (both Major and Minor) and Turnarounds. This allows us to quantify exactly how “busy” or “complex” a tune is based on the density of these cadences.

3. AI-Powered Synthesis

While the math provides the data, the narrative is synthesized using Google’s Gemini 2.5 Flash models. The AI acts as a “Jazz Harmony Analyst,” interpreting the raw data (chord frequencies, modulation patterns) to provide difficulty ratings and improvisation tips contextually.

Disclaimer

While we strive for accuracy, jazz harmony is subjective. A chord charts serves as a map, not the territory. The algorithmic analysis may interpret ambiguous harmonic functions differently than a human performer would. Use these metrics as a guide, not a rule.