What is Songle?
Revolutionary music listening experiences helped by music-understanding technologies
Songle utilizes music-understanding technologies that automatically estimate music scene descriptions and enable visualizations such as structural segments, hierarchical beat structure, melody line, and chords. For example, exploring the song structure visually enables you to discover patterns that are difficult to notice by ear. The visualized beat and chord progressions might come in handy as inspiration for your own next hit single, while seeing the vocal melody might help you hone your singing skills.
Improving recognition results by users' correction
It is not easy for humans to transcribe a melody or chords of music by just listening to it. Even trained listeners make mistakes when the music is too intricate. For our computers, too, recognition errors are unavoidable.
Our music-understanding technology provides automatic annotations for all songs to get you started. With a few intuitive mouse clicks, you can then correct the existing annotations. And because doing this on your own for a whole song may still feel a bit cumbersome, Songle accumulates everyone's contributions, including yours, to gradually refine the annotations. So even correcting only one or a few errors makes all the difference. Songle was designed to involve everyone to annotate and enjoy music together.
Amplifying your contributions.
Your contributions to Songle will not only show up as corrections of existing annotations; they will also teach our music-understanding technology to make fewer mistakes in the first place. We're planning to develop a mechanism that lets Songle learn from corrections made by users and improve its performance. Your help is very valuable to develop better music-understanding techniques.
- 2012-10-29: Keynote talk at International ACM Workshop on Crowdsourcing for Multimedia (CrowdMM 2012)
- 2012-09-26: Keynote talk at International Workshop on Search Computing
- 2012-08-29: Launch the public version of Songle
- AIST's press release on August 29th, 2012 (in Japanese)
- 2012-02-02: Launch the beta version of Songle