Thursday, March 24, 2011

Paper Reading #16: Creating Collections with Automatic Suggestions and Example-Based Refinement

Comments: Joshua Penick, Felipe Othick.

Reference Information:
Title: Creating Collections with Automatic Suggestions and Example-Based Refinement
Authors: Adrian Secord, Holger Winnemöller, Wilmot Li, Mira Dontcheva.
Venue: UIST’10, October 3-6, 2010, New York, New York, USA.

Summary: The authors of this conference paper found that current digital media management software is lacking in their implimentation of libraries, or playlists, in the case of music. Their system, rather than being a completely user generated list, or something based on tags and Boolean operators, uses natural language to crate a library that suits the user's needs. By providing input such as “some rock, no Madonna, lots of U2,”, users can get useful playlists with little work on their part.

At the end of their research, the authors found that while their computer generated playlists were fairly well regarded, users wanted a way to manually edit the items contained in their library. Their future work will delve into working with different types of mixed media and varieties of formats in a single library.

Discussion: The researcher's experiments were fairly interesting. The playlist creation seemed well suited to organizing music, but it relies heavily on information being properly labeled and tagged. I know personally my photo collection is full of IMG_009.jpg, and similar files, so such a system would not be too useful there. If I ever got to the point where I did properly tag everything, making Boolean libraries based off of their various tags would have very similar results to the computer generated library.

2 comments:

  1. I agree with you, photos are something that don't get named as much as something like music. There is some face recognition software that might be helpful but on the whole, things like photos probably would get grouped by events and time taken.

    The music programs are pretty cool that can go through iTunes and fix tracks like "Track 01" and replace it with song and artist info.

    ReplyDelete
  2. Recommender systems are pretty good, but as you mentioned user would like to have a way add or subtract from the recommend lists.

    ReplyDelete