Game Helps Computer Label Online Music

A game-powered machine-learning technique is helping computers automatically label songs on the Internet based on examples provided by unpaid music fans. University of California, San Diego (UCSD) researchers developed a Facebook-based online game called Herd It to let users submit examples of categorized music from which computers can learn. The music has labels such as “jazz,” “romantic,” or “saxophone.” The computer analyzes the music in each of these categories to find common acoustic patterns. It then uses the submitted labels on music that has similar patterns. This technology will help search engines keep pace with the great amount of music and video uploaded to the Internet and to return results for many types of music other than the most popular results. Vendors could employ the technology to create user-personalized radio services. The UCSD researchers’ long-term goal is to create a text-based multimedia search engine. They published their research in Proceedings of the National Academy of Sciences. (PhysOrg)(University of California San Diego)(Herd It)