Classifying Emotion for Information Retrieval: Three Web Sites

MoodLogic Networked Powered Music Management Client. Version 2.0. San Francisco: MoodLogic, Inc., 2001. [Requires Windows 2000, 98, 98SE, ME, or XP; Internet connectivity. Available for download from http://www.moodlogic.net. Accessed 26 August 2002. $29.95. Free trial available.]

Barnes and Noble Jazz Discovery. Developed by Savage Beast Technologies, Oakland, CA. [Available from the Barnes and Noble music page, http://music.barnesandnoble.com/jazz. Accessed 26 August 2002. Free.]

Glass Engine. Created by IBM in collaboration with Philip Glass, Watson Research Center, Hawthorne, NY. [Available from Philip Glass's Web site, http://www.philipglass.com. Accessed 26 August 2002. Free.]

"Music hath charms to soothe the savage breast, to soften rocks, or bend a knotted oak," wrote William Congreve in The Mourning Bride. Indeed, since the beginning of music history, man has recognized the emotional power of music. From Hildegard to the Affektenlehre to modern music therapy, the affective mystique of music continually inspires, fascinates, and motivates. Even if one subscribes to a formalistic view of music, it is undeniable that people of all cultures instinctively turn to music when troubled or excited, angry or content, or in need of relaxation or of motivation. The programs and Web sites reviewed here seek to exploit this emotional appeal of music while capitalizing on a widespread need to better organize, locate, and enjoy music.

Moodlogic Music Management Client

This program helps organize personal collections of digitally stored music. As its name implies, the software provides the logic for finding music to suit a specific mood.

The MoodLogic Client is freeware available for download over the Internet. It currently runs on Windows 98, ME, 2000 and XP. The program supports MPEG-1 audio files (e.g., MP3) and Windows Media files (.asf and .wma). It works together with any MP3 or media player, such as WinAmp or RealPlayer.

After a standard installation, the program can be used in a few simple steps. First, the "Add Music" function in "Activation Center" locates music files on a local computer, either automatically or manually. Songs are then sent to the MoodLogic Network, which compares these titles against a database of metadata. Given a successful match, the system returns song information including artist, song title, genre, tempo, and mood attributes. These tracks are then said to be "activated."

Tracks can also be classified manually. Users are led through a series of questions, including the song's performers, album, genre, instrumentation, whether it is live or studio recorded, and the quality of the recording, among other things. Worrisome, however, is this statement about the date of the recording: "Please enter best guess (do not leave blank)." Here, MoodLogic actually encourages users to guess a date.

After songs are profiled and activated, they can be sorted by genre, artist, tempo, and mood in the "My Active Music" area. Of particular interest is the "mood-bar," which filters songs by the emotional attributes associated with a track. Here the options are "aggressive," "upbeat," "happy," "romantic," "mellow," and "sad." Selecting one or more results in a playlist, or compilation of songs, that best expresses the corresponding mood. In a few clicks you can create a playlist of jazz songs released between 1950 and 1970 that have a slow tempo and are expressive of a "sad" emotion, for example, assuming such music is available.

Additional features include the "Get Music" button, which allows users to identify recordings similar to a given track. Clicking this opens a Web page with a list of related artists and albums. Of course, the goal here is to get users to click the "Buy" button, which leads directly to Amazon.com. Still, this feature is voluntary, not intrusive, and a nice service for those looking for related music.

The "Instant Mix" function creates a playlist of similar music based on a selected song. Two slider bars—one for genre and one for mood—allow users to dynamically increase or decrease the relatedness of tracks to one another. This leverages intrinsic associative...