Metadata: give me more, more, more, more, more
Even though I believe that metadata should more often than not be hidden from the end user, I also believe that we will need more metadata than ever and different kinds of them – attention metadata for instance.
One important technology in this respect is the automatic generation of metadata. This is something we’ve worked on for a number of years now and a topic I often discuss in my presentations. Oddly, people often tell me that they believe that automatically generating metadata is impossible. Extracting metadata from the content or generating it based on contextual clues is actually not that hard…
In the world of photography, metadata is automatically generated through for instance GPS devices that are embedded in cameras. (My cell phone includes a camera and a GPS and does this too.) Recently, face recognition is taking this a step further by detecting automatically who is appearing in your photographs. The smart trick here is to realize that most people tend “to photograph the same 25 or 30 persons, and that a camera could be taught to only recognize those specific persons”.
This is a great example of how smart tricks can often lead to solutions that are much more easy to build and much more powerful than full-blown artificial intelligence based techniques that require real breakthroughs in the state-of-the-art. I love this kind of smart solution and am sure that we can find more “tricks” of the same kind… Maybe you can suggest some additional ways to generate metadata or put them to novel uses?