Empirical Musical Cartography

This comprehensive musicology project leveraged Python and the music21 library to computationally analyze 144 classical music scores spanning the Baroque, Classical, Romantic, and Modern eras. The pipeline systematically parsed MusicXML and MIDI representations, normalizing tempo, key, and instrumentation differences before extraction.

From every score the system derived 36 distinct mathematical features, ranging from low-level descriptors like note density and interval histograms to advanced metrics such as Pitch Class Entropy and Rhythmic Density. Together these features encode a composer's stylistic DNA into a compact numerical vector.

With 36 features per score the raw data inhabits a high-dimensional space impossible to visualise directly. We applied Principal Component Analysis (PCA) to project the feature matrix onto two principal components that capture the maximum variance. The resulting 2D PCA projection map reveals a clear chronological drift: Baroque works cluster in one region, transitioning smoothly through Classical and Romantic zones, demonstrating how compositional techniques evolved over roughly three centuries.

Beyond visualisation, rigorous ANOVA testing was applied to each feature across composer groups. The results isolated statistically significant composer fingerprints, proving that quantitative models can accurately distinguish between the structural signatures of different classical composers.