August 2009: How Google beat Amazon and Ebay to the Semantic Web is a detailed and thoughtful essay on what the "Semantic Web" actually means and how it may work, written as a 2009 retrospective of Google's dominance of the medium.
I suspect that the security people are having vicious nightmares about this sort of thing; Paul Ford's essay doesn't mention security at all (presumably it's implied) but the notion that a web of trust is built using RDF has severe security implications when we consider how common website defacements are today. That takes on a whole new layer of importance if you can change the bank's RDF output to state that you are in credit, rather than just writing "rEnEgAdE KhEmIsT is l33t!" on someone's page.
I'm also not seeing how this solves the perennial search problem;
knowing the data inputter's mind to guess how they described an object.
Right now, if you want to search for, say, a VW Microbus, do you search
for "VW Microbus" as a phrase? "VW" and "Microbus" as separate words?
What about "Volkswagen"? And "Type 2", another name as which it was
known? There's no overriding classification system. Now, possibly,
extending this RDF-driven semantic web metaphor, we could have:
[Volkswagen plc] says that (Microbus) is also called (Type 2)
[Volkswagen plc] says that (Volkswagen plc) is also called (VW)
but, again, is this basic classification effort going to be put in? This reminds me a little of AI research, where one possible angle of attack is to attempt to provide a computer with axioms of common sense, a list of facts about the world. So rainbows are made by light split by rain. Wood floats and stones sink. Dreams aren't real. Run away from wild lions. An old issue of Dragon Magazine recommended that, when designing worlds, you think of taking an alien around this world. The alien speaks perfect English (or your native language) but doesn't understand anything about our world. So you start to explain; cars run on petrol, most people are right-handed, thinking about bad things doesn't make them happen. At the end of the day you'll realise that you've barely scratched the surface of what you know. All this data will need to be represented in the semantic web in order to make it truly useful, I believe, and no-one will do so, because no-one owns it -- VW might fill in data about their Microbus, but who owns the notion that dreams aren't real? Or that most people are right handed? More to the point, some people think that dreams are real; where does the notion of conflicting information fit into the semantic web? Search engines are not good at handling this sort of data; how does the amazing RDF index work if some of the information conflicts? As with knowledge management, the way computers currently handle this is to give their human users all the data, conflicting or not, and make them sort the inconsistencies out.
Nonetheless, this is a fascinating essay describing a fascinating concept. How can we make it happen? It doesn't really work until everyone's using it; like a lot of technology, especially technology relating to the internet and communications, a small version of it is pretty useless, and the gains come when everyone uses it as a whole or lots of small pockets all join up. The "small pockets" approach is better, if a small pocket can exist and be useful; I can see, for example, a university campus working on this sort of RDF scheme for buying and selling products a la the postulated Google Marketplace. With lots of these small islands existing, the infrastructure for a massive global network is in place; this is how the very Internet got started, after all.
Keep your eyes peeled. And buy Google shares now.
© Aquarius, July 2002