Latent Evolutionary Signatures: A General Framework for Analyzing Music and Cultural Evolution
Bibliographic record
Abstract
Latent Evolutionary Signatures: A General Framework for Analyzing Music and Cultural Evolution © 2024 Jonathan Warrella,b,1, Leonidas Salichosa,b,e,1, Michael Ganczc,1, Mark B. Gersteina,b,d 1 equal contribution a Program in Computational Biology and Bioinformatics, Yale University, New Haven, CT 06520, USA. b Department of Molecular Biophysics and Biochemistry, Yale University, New Haven, CT 06520, USA. c Department of Music, Yale University, New Haven, CT 06520, USA. d Department of Computer Science, Yale University, New Haven, CT 06520, USA. e Department of Biological and Chemical Sciences, New York Institute of Technology, New York, NY 10023, USA. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ This repo contains optimal harmony and form+harmony based models for analyzing popular music as an evolutionary structure. Our model architecture is based on a traditional VAE, but with an energy-based prior that penalizes a measure of ‘evolutionary distance’, in this case informed by temporal distance across song release dates (see schematic below), in the latent space. The key output of each model is a set of latent ‘evolutionary signatures’, or characteristic distributions of chord/form k-mers that can be used to predict the date and genre of each song. We use the McGill Billboard corpus of popular song annotations as our database. The subdirectory ‘harmonic_formal’ contains the code for training the model based both on chord progressions and formal features. The subdirectory ‘km4’ contains the code for training the model based on chord progressions of length 4. In addition to the optimal models, we include some other configurations that we tested, including variants on the coarse-graining of formal units (models suffixed with ‘_A’, ‘_B’, and ‘_None’, referring to projection matrices A and B, where A retains formal categories that comprise the first 99% of the data, and collapses all others into a category of ‘other’; and B bins together semantically similar formal units), and different means of normalizing formal feature vector ‘X_struct,’ which in its raw form contains the counts of each formal category (models suffixed with ‘_binarized’ or ‘_zscore’ utilize these normalizations. These are contained within the ‘models’ folder of each subdirectory. Additional items of interest, including code for processing the McGill Billboard dataset, are included in the ‘supplemental’ folder. The ‘McGill-Billboard’ folder contains all raw data. All code is currently written for MatLab.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.003 | 0.012 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.004 | 0.004 |
| Science and technology studies | 0.001 | 0.002 |
| Scholarly communication | 0.003 | 0.004 |
| Open science | 0.003 | 0.003 |
| Research integrity | 0.002 | 0.003 |
| Insufficient payload (model declined to judge) | 0.005 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".