Re: [Rhythmbox-devel] Re; Ratings feedback



On Mon, 2004-02-16 at 10:12, in7y118@public.uni-hamburg.de wrote:
> Since we're doing user profiles here:
> 
> About me:
> - I have a collection of 5000 songs.
> - Rhythmbox always plays the whole library, repeat and shuffle activated.
> - I never rate a sigle song, nor do I think I'll ever do it.
> - I have bound the F12 key to "rhythmbox --next" in metacitys keybindings. 
> That way I can go to the next song from anywhere.
> 
> About my usage of Rb:
> - I start up Rhythmbox, then press F12, so it starts playing.
> - Whenever there's a song I don't like, I press F12. That happens quite often. 
> (I skip 2 out of 3 songs at least)
> - Occasionally (every other day or so) I think "I want to hear XYZ now". In 
> that case I switch to Rb, enter XYZ into the search field, double click the 
> song to start it, then clear the search field again, so it continues with 
> shuffle.
> 
> What this behaviour should mean to Rb:
> - Listening a song to the end means I liked listening to it. I'd probably like 
> listening to it again.
> - Skipping it means I don't want to hear it right now. Maybe later, but most 
> likely not "soon".
> - Selecting a song explicitly means I like it a _lot_. I probably wouldn't 
> mind if it was played three times in a row.
> - Adding a song to the library means I like it. I wouldn't add it if I didn't 
> want it to be played.
> 
> What I want Rb to do:
> - Make it so I never need to use F12.
> 
> Benjamin

What you describe is surprisingly similar to the algorithm used in RB
right now.

If you 'jump' to a song by manually selecting it, the rating will go up
more.
If you listen to a song all the way through and you have been active on
your computer since the song started, the rating will go up more.
If you skip a song before the last 30 seconds, the rating will go down.

It is a bit more complicated than that, but you get the gist of it. 

You are probably the exact use case the algorithm was designed for; it
also works for me well.  This is the same way that IMMS works, and that
seems to be popular.  So, I doubt the algorithm has any huge flaws.

The one problem mentioned was concern that it would not adjust the
ratings of songs quickly enough.  I personally think that slowly rating
songs is good, because then you get less jittery ratings when your
listening mood changes from session to session.  But, I again defer to
IMMS here, since when I wrote the auto-rating, I merely took the amount
IMMS scaled songs and translated it for RB's 5 star system.

-mt

This is a digitally signed message part



[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]