Metadata and relationships
Sorry... two posts prompted by Scott Wilson in quick succession, but I couldn't help myself. Scott writes about metadata here, in response to Wayne Hodgins' post on the Future of Metadata and I have to say I agree with every word.
I'm not sure that I buy Wayne's emphasis on automated metadata generation. In many cases, if I can automatically generate a bit of metadata and give it to you then the chances are that you can just as easily generate it yourself. I know that isn't always the case, because in some cases my context, capability and/or knowledge might allow me to generate something that you can't, but I bet it is true in a lot of cases.
More fundamentally, and I think as Scott is arguing to a certain extent, why generate a bit of metadata for someone else to index at all, when they can just as easily index the object itself. OK, at the moment that really only works for textual resources... but surely that will change over time?
However, it seems to me that one of the key things we can still usefully capture in metadata is the relationships between stuff. This is a version of that... this is derived from that... and so on. And, often, it is an intellectual exercise to work out how two things are related. Sure, on the Web we can index the full text of everything and we know that things are related to other things in a general sense (courtesy of the hypertext link), and we can build very powerful systems on the back of that knowledge (e.g. Google). But we don't know anything in detail about how things are related to each other - and that is very limiting for some applications.
Comments