Context as Metadata

Context - (c) Jeremy Noble More than a year ago, Henry Story blogged about Keeping track of Context in Life and on the Web. It is about the context of the story you’re telling, as essential background information for the general audience and distracting bloat for the initiated at the same time.

The conclusion is that, using a semantic web approach, you could provide links to as many contextual facts as you like, without the need of directly exposing these to the observing end user. Just use those links for queries and matching algorithms wherever appropriate.

In other words: don’t bug me with redundant metadata if I don’t need it. This might be even more true for content creation: just read Cory Doctorow’s Metacrap article again and you know why.

Years ago, almost immediately after I bought my first digital photo camera, I started to realize why metadata is important. In a few words: taking pictures is easy, storage space is cheap and deleting images is a pain. You need to carefully compare and make sure to pick the best one. So, hundreds, soon thousands of images started to pile up in the form of un-imaginatively named blobs, like “IMG_1123.JPG”. Essentially, these images get lost as the proverbial needle in a haystack.

Now you could put all those images in folders, labeled after an event, date, person or whatever. But this is a tedious job and only provides a very flat view (you don’t even want to think about creating nested or linked structures on your file system).

Then, I soon found out that every digicam image has embedded EXIF meta data, which proved to be of huge value for tracing back those lost images. If I know that a shot was made during some event, I only need to look up the events’ date and browse all images shot during that period.

Then iPhoto came around, with the possibility to add tags (with a terrible interface, use Keyword Assistant instead!), ratings and multiple album folders. Providing even more metadata and control to find your images at a later time.

There’s just one problem left: entering and assigning all that meta data by hand is still much work if you have hundreds of images to go. Errors are quickly made and hard to detect when you’re focused on other things, such as composition and image quality.

So, what I really need is as many context facts as I can automatically gather. For instance, having GPS location data is invaluable if I have no idea about the date I’m looking for, or to select the best image from over many years. Need a sunset sky reflected in a window of your home and you know that you might have a photo around? Good luck without location data.

I could add more: an outside shot with freezing ambient temperature gives a much improved chance to find that icy skating scene some winters ago. Best of all, there are numerous applications for combining simple, trivial meta data facts that I now can not even imagine I will ever need. But if the time comes that I’m looking for that very specific image, I’ll be glad that I have all that context data stored.

Linking things, facts and events on unimagined context attributes will provide real value, which a rich data web will provide. Just start adding context just because we can, the applications will follow rather sooner than later.

Standardized semantic web tools and technologies provide a sound basis for this whole idea. For example, use RDF triples for all those context metadata. Then a SPARQL query can be used to hunt for any data you might get interested in later on.