Post from Jean Thilmany:
News from engineers at the University of California San Diego suggests that a computer algorithm they’ve developed and are calling game-powered machine learning, would enable music lovers to one day search every song on the web well beyond popular hits, with a simple text search using key words like “funky” or, even more precise, “spooky electronica.”
The engineers found that a computer can be taught to automatically label every song on the Internet using sets of examples provided by unpaid music fans. And the results are as accurate as using paid music experts to provide the examples, saving time and money, said Gert Lanckriet, a professor of electrical engineering at the UC San Diego Jacobs School of Engineering, who led the work. The results were published in the April 24 issue of the Proceedings of the National Academy of Sciences.
The eventual hope is to create a text-based multimedia search engine that will make it much easier to access the explosion of multimedia content online. Today, humans working round the clock labeling songs with descriptive text could never keep up with the volume of content being uploaded to the Internet, Lanckriet said. For example, YouTube users upload 60 hours of video content per minute, according to that company.
Lanckriet foresees a time when, thanks to this massive database of cataloged music, cell phone sensors will track the activities and moods of individual cell phone users and use that data to provide a personalized radio service—the kind that matches music to one’s activity and mood, without repeating the same songs over and over again.
Speaking for myself, I’ve never had much luck with Pandora Radio, which purports to learn your musical preferences based on your input of the type of songs that you enjoy listening to. You’re then prompted to give each song Pandora subsequently plays a thumbs up or down, and the radio station, backed by its algorithm, goes on to hone in on your musical taste and refine the songs it plays for you.
I know the UC engineers said their machine-learning tool will go well beyond Pandora and the “if you like” style algorithms, but my tastes are broad, from rag time to hip hop, and such algorithms seem to have a hard time accounting for that—and I want to find my way to my own songs. For some reason, the “if you like this, then you’re sure to like this” style of finding my way to new songs and musical genres has never appealed to me. I’d rather be an explorer on my own.
“What I would like long-term is just one single radio station that starts in the morning and it adapts to you throughout the day,” Lanckriet has said. “By that I mean the user doesn’t have to tell the system, ‘Hey, it’s afternoon now, I prefer to listen to hip hop in the afternoon. The system knows because it has learned the cell phone user’s preferences.’”
But I can’t be alone when I say I may enjoy listening to hip hop for an hour or so some afternoons, but then want to hear three hours straight of Dizzy Gellespie on another day. Bottom line is music, like taste in food, is highly personal and people tend to eat what they have a craving for right then, which varies by day and by hour.
There are just some human preferences a computer can’t predict or follow.