Added: 2 years ago
From: JacobsSchoolNews
Views: 1,909
Sort by time | Sort by thread (beta)

Link to this comment:

Share to:

All Comments (4)

Sign In or Sign Up now to post a comment!
  • I wondered about this classification too. How does the algorithm deal with songs with multiple movements like The End to a Sad Film? The first part (more upbeat) might be classified as waking up music, but the later parts of the song are tragic like Vigilant described.

  • I call this "the Bohemian Rhapsody problem"! 

    We have an algorithm that can automatically divide a song up into homogenous segments like chorus / verse / bridge etc. Then we tag each segment separately. The trick is how to combine these segments together? Do you just add / average the description for the chorus and the verse? Or do tell listeners "the verse of this song matches your query for a mellow piano ballad " (but then, in the verse, crazy guitars kick in!)

  • Yeah, Pandora would be a great source of data to train the algorithm on. Pandora could also benefit a LOT from a computer that could automatically add new songs to their database - right now it takes about 30mins for a human to listen to the music and annotate it before Pandora can add a song. 

    Means that new / unknown bands are never going to make it...

  • I am not used to waking up to the sound of a sad film. Why did the algorithm decide to wake up to Nickolaus Mueller's tragic song? I'd rather rather wake up to "Oversleeping" by I'm From Barcelona or "Up, Up, and Away" by KiD CuDi.

    Five stars though! I fully support this research. Ideally Pandora would fund this research. There would be tremendous synergy from having a professionally tagged database to train your algorithm on.

Loading...
Alert icon
0 / 00Unsaved Playlist Return to active list
    1. Your queue is empty. Add videos to your queue using this button:
      or sign in to load a different list.
    Loading...Loading...Saving...
    • Clear all videos from this list
    • Learn more