MoodLogic
MoodLogic was first conceived in early 1999 by Tom Sulzer when he was struggling to manage his own CD collection. After discussions with focus groups and development with friends at MIT, he concluded that an inventive means of recommending music would be a viable business. So Sulzer and his friends started MoodLogic in October 1999. However, the site is not alone in its crusade. MoodLogic takes advantage of a companion site called Jaboom.com, which is designed to collect music data.

Brett May, MoodLogic COO

MoodLogic's interface introduces an interesting way to think about music. It has categorized songs with conventional parameters like tempo, era, and genre and has introduced the music-rating concept of mood. By using "mood magnets," you can determine which songs best represent your search. The magnets lets you control eight moods: aggressive, upbeat, happy, romantic, sentimental, mellow, sad, and brooding. Brett May, chief operating officer of MoodLogic, says, "The magnets are a great way to visualize a song. Every song occupies a unique position in a 100-dimensional universe. These magnets provide you with a number of axes along which you can define music. For example, you can choose the genre and tempo, and then you organize the songs around an interesting feature called 'mood.' So as you drag the mood magnets around, the songs will gravitate toward what best describes them."

The MoodLogic Magnets

Behind the magnet interface, though, is a technology fueled by the opinions and considerations of avid music listeners. May describes the music data collection process as a form of collaborative filtering. "We have a site called Jaboom.com, where folks go and rate music. When you visit, people are asked to fill out a 30-question survey. Under normal circumstances, we collect a variable number of votes for these questions. So we'll collect, say, 150 votes on whether or not a song is happy. Then we have statisticians here who determine how many votes we need before we know we have a correct and representative answer. Finally, we have a ring of experts in-house, and we have them look at the data and remove the answers that might be anomalies. Once enough votes on a specific song are in, we can generate a score and then enter that information into the database."

Currently, MoodLogic has some 200,000 songs in its database, though it has information on 500,000. You can't actually download the music off its site; rather, it uses the database for recommendations only. The MoodLogic staff is in the process of releasing songs into the magnet system at a rate of "thousands per week." As the songs are analyzed and defined, they will ultimately be entered into the database. Truly, a gargantuan task that's appropriate for only the most deliberate of music aficionados.

Finally, MoodLogic plans to implement personalized search technology early next year. May concludes, "That's the next release: We know who you are, we know what you like, and we know what your prior searches were. We can use that data to assign greater or lesser weights to a song's probability of being in your search. That's the next release toward better collaborative filtering." May believes that MoodLogic will have 500,000 songs ready in three or four months, just in time for the personalized search functionality to roll out. Happy New Year!

A sonic approach to music discovery