Commuting again, an opportunity to listen to some music. I recently bought a Mac to use as my home machine, and one of the unanticipated benefits has been iTunes. iTunes has a music recommendation system called Genius. Click a song and Genius can create a playlist of your own stuff and recommend songs that you might want to purchase. I have found it to be almost ridiculously good, matching the tempo and mood of the songs that I give it. I assume that Genius is based on the kind of social recommendation systems that companies like Amazon have nearly perfected: people who bought that, also bought this; people who listen to that song tend to put it in a playlist with x, y and z. One thing I have noticed, however, is that Genius sometimes doesn’t like to jump genre boundaries. It will put hip-hop tracks with electronica, but doesn’t seem to like to mix disco and funk with remixed disco and funk, or jazz with remixed jazz, or funk with funky jazz. Does this tell me something about the algorithm or something about peoples’ genre preferences? I’ve long thought that humanists could use a similar recommendation system for text mining: people who consulted / cited that source also looked at these other ones. I’m pretty sure this isn’t what Walter Pater had in mind, but we could stand to learn something from the condition of music.
Tags: AudioCulture, DDH-Evening, DDH-Reflecting, DDH-Travle, Research, SocialRecommendationSystems, TextMining