Commentary by Mark Wahl, CISA
Metadata in Personal Content Experience (20080128)
Historically, the idea of managing "metadata" in computer systems has primarily been of interest to practicioners of data warehousing or professional multimedia content producers. Recently, the book Personal Content Experience: Managing Digital Life in the Mobile Age by Juha Lehikoinen, Antti Aaltonen, Pertti Huuskonen and Ilkka Salminen, published in 2007 as paperback and Adobe, argued for the importance of a broader understanding of the value which metadata provides to emerging applications: 2 of the 8 chapters in a book on mobile multimedia application have "metadata" in the title.
In one motivating scenario they present, an artist has been taking hundreds of pictures with a digital camera, and frequently uploads her pictures to a photo sharing web site. While uploading, she adds descriptive labels and notes to her photos as they're going onto that site. Later, the company that provided the photo sharing web site ceases business. She still has all of her photos stored in her computer, but the labels and notes are gone and she has no way of searching her photos: it wasn't clear to her that the metadata she was entering wasn't going to be part of her copy of the image files. It's unlikely that she'll go through the process of re-tagging all of the images, and now they're just wasted space: directories full of files with arbitrary names IMG_nnnn.jpg without any context.
The authors suggest that the traditional models for categorizing and administering metadata have a 'library sciences' flavor that doesn't mesh with the demographics for use of today's media in need of metadata. The iPod-toting youngsters
are looking for audio files indexed by metadata attributes such as Genre, Energy, Hipness, not Compression algorithm or Rightsholder.
Furthermore, a key problem in their environment, smart phones and other mobile devices, is that the likelihood of someone manually entering any significant amount of metadata while they're importing content (downloading a song, taking a picture) to the device is basically nil, since they user is likely to be (a) on the move, (b) using a device with a greatly reduced keyboard, and (c) not interested or motivated to do so. However, many existing implementations for automated metadata extraction have been oriented to details not of interest to this category of user. The authors suggest that in particular relational categories of metadata, which link two or more objects together, and the context of interaction have not been fully appreciated.