Tuesday, February 15, 2011

Paper Reading #9 - Creating Collections

Comments:
Comment 1
Comment 2

References:
Title: Creating collections with automatic suggestions and example-based refinement
Authors: Adrian Secord, Holger Winnemoeller, Wilmot Li, Mira Dontcheva
Venue: UIST 2010, Oct 3-6, 2010

Summary:
In this paper, the authors describe a new hybrid method of creating playlists that merges automatic and manual creation methods; they then show two programs they created for managing media using this method, SongSelect and PhotoSelect. They begin by talking about automatic playlist making methods like iTunes' Genius tool and how many users thought it didn't give them what they want. Then, they discussed how many users selected music for their playlist that sufficed instead of looking for a best candidate despite having concrete goals, a method they call satisficing. So, when they designed their two applications, they kept a two step process: keywords are entered into a box that automatically generates a set of songs or pictures meeting the criteria; then, the user selects the best media for the list. Furthermore, the user can get further suggestions from the list if necessary to further customize their choices.

Discussion:
I thought that this paper was interesting because it's trying to solve an issue that I run into often. As they mention in the article, automatic playlist makers often give unsatisfying results, and doing it manually is a giant chore. I actually hope that this idea takes off.

However, and this is really picky of me to mention, I didn't like the use of a portmanteau inside of a professional paper. I feel like it looks unprofessional, and they could have still gotten their point across without it.

(image courtesy of: megaleecher.net)

6 comments:

  1. I think Itunes' album coverflow is the way I prefer to display the music I have. I prefer seeing all the music I have rather than in a playlist.

    ReplyDelete
  2. I really liked this article as well. It actually presented a problem as well as a solution, which most of my readings have not. To the above comment, this system would have all the music available to choose from, but the interface is designed to create playlists from the entire library. It was my understanding that you could still play from the entire library as well.

    ReplyDelete
  3. I love the genius feature of iTunes, so I think this would be a very welcome improvement. It would prove very useful if implemented.

    ReplyDelete
  4. One problem I see with all this is a huge increase in metadata. Either the metadata stays with the file or the application, but if your like the guy with 70,000 songs...well, thats just ridiculous. But hey, our songs are like 5-6 MB each, so what's another 10% on top of that. Honestly, i think it's time for a new compressed music file format with relatively high fidelity.

    ReplyDelete
  5. Interestingly enough, the metadata seems like a lot for lossy compression algorithms, but if you're like me and use a lossless algorithm like FLAC, it becomes an extremely small fraction of the file size.

    As far as metadata is concerned, there needs to be a source for complete metadata for all of this music first, a majority of tracks I own are missing a lot of data that might be useful for a program like this.

    ReplyDelete
  6. I run into this problem a lot however it does take many hours to be able to organize large amounts of music. I have over 10,000 songs and many times don't listen to songs because I don’t have a good playlist. I think however it is impossible to make a playlist of songs automatically that a user will like because of everyone’s personal preferences.

    ReplyDelete