Closing Keynote: Paul Lamere
summary
Forty minutes of concrete results: 14 experiments on Spotify-scale data that mine listener plays, cultural text, metadata and acoustic features to build apps and metrics. Highlights: the Aerosmith anomaly tying streaming spikes to news events; playlist recipes from 2 billion playlists and demographic filters; a passion index (In Flames averages ~115 plays/fan); scrubbing-based hot‑spot detection, a Python/JS remix library, The Infinite Jukebox, the Bonhamizer, an Autocanonizer and a live Girl Talk in a Box demo — all prototypes that run and often ship.
AI-written summary · about
chapters (24)
- 0:00Welcome to RailsConf
- 1:03From trombone to music tech
- 3:13Music hacking with Spotify data
- 4:16Aerosmith Anomaly in music data
- 9:36Walkman to 30M songs
- 11:22Music metadata challenges
- 14:33Band names getting longer
- 16:42Cultural data and artist similarity
- 19:51Boil the Frog playlist generator
- 22:19Listener data and passionate fans
- 26:33Mining 2B playlists for context
- 29:00Playlists for demographics
- 32:41Best parts of a song with scrubbing
- 38:17Acoustic data and interactive listening
- 39:43In Search of the Click Track
- 41:24Finding dramatic moments
- 42:46Visualizing music with 3JS
- 45:38Automatically remixing music
- 47:31The Bonhamizer and John Bonham
- 49:18Swinger and making songs swing
- 50:00Infinite Jukebox
- 52:06Autocanonizer
- 54:18Girl Talk live performance
- 58:14Love your data and hack
related talks
28mLearning Fast: Home Experimentation and Open Research at Spotify
20mRediscovering music
13mRevolutionizing the Music Industry with Data Science
50mBetter Spotify Playlists through Music Theory & Discrete Optimization
40mPowering Spotify's Audio Personalization Platform
10mKeynote with Ben Fry
24mHow Spotify built an audience framework that lasts
18mIt's a Bloody Mess! Solving One of Music’s Biggest Challenges