How ethannbrown/soundstages works
soundstages is a research pipeline that collects music metadata and audio features, cleans and joins them into an analysis dataset, evaluates popularity models across artists and quantiles, and produces findings figures. Its main workflow runs from artist discovery through track-feature collection, preparation, modeling, evaluation, and reporting. The graph retains the documented collection and modeling branches, including both linear and quantile/XGBoost analyses. The README establishes the stage sequence; source excerpts and the index support selected internal calls. Unexcerpted implementation details are not inferred.
Main entry points: artists.py Artist discovery · prep.py Model table and folds · folds.py Fold utilities · clean_tracks.py Track cleaning
Components 27
Data collection
- Artist discovery
src/collect/artists.py - Listener counts
src/collect/listeners.py - Genre resolution
src/collect/resolve.py - Genre mapping
src/collect/genres.py - Genre rules
src/collect/rule.py - Track ID resolution
src/collect/ids.py - Album collection
src/collect/albums.py - Track collection
src/collect/tracks.py - Audio features
src/collect/features.py
Dataset preparation
- Artist merge
src/prepare/merge.py - Album cleaning
src/prepare/clean_albums.py - Track cleaning
src/prepare/clean_tracks.py - Text normalization
src/prepare/text.py - Dataset build
src/prepare/build.py
Modeling
- Model table and folds
src/model/prep.py - Fold utilities
src/model/folds.py - Baselines
src/model/baselines.py - Linear models
src/model/linear.py - Quantile regression
src/model/quantile.py - Era quantiles
src/model/quantile_era.py - Quantile XGBoost
src/model/boost.py
Evaluation and reporting
- Prediction evaluation
src/model/evaluate.py - Results figures
src/report/figures.py
Other components
- Researcher
- Every Noise
- Spotify pages
- ReccoBeats API
Connections 42
- Every Noise → Artist discovery: provides artists
README.md - Spotify pages → Artist discovery: provides artists
README.md - Artist discovery → Listener counts: collects listeners
README.md - Artist discovery → Genre resolution: resolves genres
README.md - Artist discovery → Genre rules: uses rules
src/collect/artists.py - Artist discovery → Genre mapping: uses mapping
src/collect/artists.py - Genre resolution → Genre mapping: uses mapping
src/collect/resolve.py - Artist discovery → Artist merge: merges dataset
README.md - Listener counts → Artist merge: merges dataset
README.md - Genre resolution → Artist merge: merges dataset
README.md - Artist merge → Track ID resolution: resolves IDs
README.md - ReccoBeats API → Track ID resolution: resolves IDs
README.md - Track ID resolution → Album collection: collects albums
README.md - ReccoBeats API → Album collection: serves albums
README.md - Album collection → Album cleaning: cleans albums
README.md - Album cleaning → Track collection: collects tracks
README.md - ReccoBeats API → Track collection: serves tracks
README.md - Track collection → Track cleaning: cleans tracks
README.md - Track cleaning → Text normalization: uses text rules
src/prepare/clean_tracks.py - Track cleaning → Album cleaning: reads album years
src/prepare/clean_tracks.py - Track cleaning → Audio features: collects features
README.md - ReccoBeats API → Audio features: serves features
README.md - Album cleaning → Dataset build: joins dataset
README.md - Track cleaning → Dataset build: joins dataset
README.md - Audio features → Dataset build: joins features
README.md - Dataset build → Model table and folds: prepares model table
README.md - Model table and folds → Fold utilities: assigns folds
src/model/prep.py - Model table and folds → Baselines: fits baselines
README.md - Model table and folds → Linear models: fits regressions
README.md - Model table and folds → Quantile regression: fits quantiles
README.md - Model table and folds → Era quantiles: fits by era
README.md - Model table and folds → Quantile XGBoost: fits boosted models
README.md - Baselines → Fold utilities: uses folds
src/model/baselines.py - Linear models → Fold utilities: uses folds
src/model/linear.py - Quantile regression → Fold utilities: uses folds
src/model/quantile.py - Era quantiles → Fold utilities: uses folds
src/model/quantile_era.py - Quantile XGBoost → Fold utilities: uses folds
src/model/boost.py - Baselines → Prediction evaluation: writes predictions
README.md - Linear models → Prediction evaluation: writes predictions
README.md - Prediction evaluation → Results figures: reports scores
README.md - Quantile XGBoost → Results figures: writes score and SHAP results
src/report/figures.py - Researcher → Artist discovery: runs pipeline
README.md