Research Dashboard
How does the economy
shape what we listen to?
Top-charting songs from the US Billboard Hot 100 and UK Official Charts, analyzed through computed lyric embeddings and audio features — then correlated against macroeconomic indicators. No subjective labels, only vectors and math.
—
Songs analyzed
—
Years covered
—
Natural clusters
—
Significant correlations
Methodology Pipeline
01
Collect
Billboard + UK charts
Lyrics via Genius
Audio via Deezer
Lyrics via Genius
Audio via Deezer
02
Embed
384-dim lyrics (MiniLM)
33-dim audio (librosa)
417-dim combined
33-dim audio (librosa)
417-dim combined
03
Reduce
UMAP 417D → 3D
Cosine metric
15 neighbors
Cosine metric
15 neighbors
04
Analyze
HDBSCAN clustering
Pearson correlations
Temporal drift
Pearson correlations
Temporal drift
3D Vector Space Explorer
Interactive Plotly visualizations — rotate, zoom, hover for details.
Economic Correlations
Pearson r between year centroid positions (X/Y/Z) and economic indicators.
Cumulative Musical Drift from 2006 Baseline
Year-to-Year Drift Magnitude
Sonic Diversity per Year (Mean Distance)
All Correlation Results
| Indicator | Dimension | Correlation (r) | P-value | Significance |
|---|
Natural Clusters
HDBSCAN clustering with no predefined k — pure unsupervised grouping.
Cultural Outliers
Songs furthest from their year's centroid — they broke the mold of their era.
| # | Song | Artist | Year | Chart | Distance from Centroid |
|---|