Evolution of Music, Cantometrics
summary
Claims that digital cantometrics recovers historical signal from performance features using a cleaned Lomax dataset of 5,000 songs (800 cultures, 44 variables) and a tailored K‑means clustering algorithm that produces eleven cantographic clusters. Software-driven clustering plus human review finds geographically coherent groupings, a 13,000-year signal linking Patagonian, Inuit and Ainu vocal styles, and argues musical styles are unusually conserved (deeper than language). Workflow, scaling to ~100,000 songs, and copyright impediments to open databases are discussed.
AI-written summary · about
chapters (9)
- 0:00Music and evolution
- 2:06Introducing cantometrics
- 5:24Reconstructing musical history
- 8:41Reviving the data
- 12:28Bridging humanities and science
- 15:06Why humans make music
- 17:25Geography and migration patterns
- 22:53Speed of musical evolution
- 26:33Copyright and future research
related talks
22mSong Recommendation Using Fourier
05mData science and streaming music - Interview with Jonathan Gosier
02mis everywhere
54mKeynote: Perfectly Imperfect: Music, Math and the Keyboard
34mBuilding a Large Music Recommender Leveraging AI, Deep Learning and Human Expertise
04mMusic, the window into your soul
03mCombining creativity and analytics
02mTamara Kolda, Distinguished Member of Technical Staff, Sandia National Laboratories @ MLconf SF Part