Blog: The meaning of the meaning of meaning

My recent piece on search engines provoked this email comment from Tom Welsh (Reg Dev contributor and Senior Consultant with Cutter Consortium):

"Most people are quite unaware of the yawning gap between data and "knowledge", which is what semantics is all about. As for the term "unstructured", I concluded long ago that it is thoroughly subjective, and means whatever the given speaker finds unclear."

Tom's quite right and it set me thinking. I remember when Microsoft's XML guru told me that XML has all the semantics anyone needs; of course, by itself it has no semantics at all (try reading a piece of valid well-formed XML in Mandarin and writing a program to process it correctly) which is why we need ontologies and the semantic web and so on. Our XML guru admitted this point, once he'd thought about it a bit.

And that's the issue. Issues of Semantics are "obvious" until we think about them.

We recognise an XML invoice - in English - but if we try to process it without knowing the legal significance of some of the terms we may come a cropper. Sometimes our assumptions will be right, however, and we'll get lazy - but assumptions about semantics are still assumptions.

I know that a customer is someone we've done business with, who has a validated credit rating; but you may think of a customer as anyone who has expressed an interest in our product. Mix well-formed XML documents corresponding to both types of customer in one database and I think we may have business-level issues to resolve sooner or later - but the associated automated processing will probably work just fine, up to the point of delivering the wrong answers.

I did think of taking a three-way view in my piece - Data is just raw data; Information (data plus metadata, schema, semantics etc) is data that we understand; Knowledge is information that we can, and do, actually use for something useful. Now, does semantics come in with Information or Knowledge? It's a question of semantics (and, yes, that's a different link and includes a slightly different meaning).

But it gets messier. Even raw data can be knowledge, perhaps, if we use it for something; but it's high risk knowledge because if we are using data blindly, there's a chance that we'll use it inappropriately...

Besides, is such raw data really raw data and not just low-value information? If we know that when a figure 6 appears on the dial we must splifflicate the bifurcator (even if we nothing more about what this means), that "6" data item has become both information (we know it relates to bifurcators) and knowledge (we use it), Which gets us into knowledge quality metrics and the like...

Just don't go there....

But someone will have to...®