If you’ve read this, then you know all about how Shazam, the sound-identifying app, works on a technical level using algorithms. But how do they ensure users can ID all songs, not just whatever’s the hit that week on the radio?
The New York Times has a really nice piece on the hunter-gatherers who go out collecting music for tagging. Only a small percentage of the tracks the app is capable of recognising actually come from record labels, so it’s up to employees like Charles Slomovitz to source music and feed it to their database.
From there, the fingerprints of a song are uploaded to a database, and then matched by users who search for that desired ditty. While Shazam hasn’t released figures of how many songs they can’t identify (I’m willing to bet it’s a good amount, based on my efforts with the app alone), there’s no denying that they’ve got the popular songs down-pat. It’s the indie stuff they’re working hard on, with people like Slomovitz actually going out onto the street or listening to local college radio stations to find underground music, plus scouring Hype Machine and all the charts for new songs.
It’s an enviable job for sure, but I can’t help but think they must feel like snow-shovellers (or hamsters in a wheel), always digging but never clearing the mess. [New York Times]