UFO's seen growing on the Web

UFO's, or Unified Foundational Ontologies, have struck again on the Semantic Web mailing list last week. Apparently they do this quite regularly, but the pattern of the attacks is not quite clear. They arrive in groups it seems and attack with incredibly long posts making references to huge numbers of philosophical and logical arguments from well known and respected accademics. Clearly the typing fire power of these aliens is way superior to mine or any I have seen in real life. Furthermore their reading capacity must be astounding, as they will throw out book references and expect one to have read them before the reply is over. But they have a weakness which is, it seems, rooted in biology. On the surface though they make the following claims:

  1. There are some structures that RDF is missing that makes it impossible for it to express some/many very important UFO concepts.
  2. That previous experience has led to disaster (known as the AI winter), because they did not work with the right UFO
  3. There are no UFOs yet, so no work can be done

Now on point 1. it is quite clear that the Aliens don't have a very good grasp of RDF as I see it. They are still amazed that RDF is not about syntax. You can write RDF in a huge number of different ways, to expressing the exact same thing. The reason this is possible is because RDF is built on the relation between names as URIs and the things they refer to. This relation is also known as Semantics. Since the names used are Universally unique (URIs), it does not matter much if your write your sentences aRb or Rab or abR or ≤R sub="a" obj="b"≥ or however you prefer. This emphasis on universally unique names is completely new. It changes things fundamentally, and yet it seems so innocuous.
In any case it is very dubious that there are things that cannot be expressed in rdf (especially with the addition of graphs as found in SPARQL) but that could be expressed in other ways, as [check reference] it is at least as powerful as first order mathematical logic.
Finally there is some existential proof to the contrary with DOLCE, "Descriptive Ontology for Linguistic and Cognitive Engineering", for which there is an OWL version.

Now on point 2. the case has been made that there may never be one UFO design that is clearly better than the others, and even a lot of skepticism that UFOs exist or are even possible. Harry Halpin provided some extensive references to back up that claim, and neatly wrote this out in a very interesting article The Semantic Web: The Origins of Artificial Intelligence Redux. (Well from the lengths of Harry Halpin's posts I sometimes wonder if he is not one of the Aliens too). But of course one only needs to look at Guha's thesis Contexts: A Formalization and Some Applications (summarised in my previous post Keeping track of Context in Life and on the Web) to understand that creators of the Semantic Web were very aware and had a very clear understanding of the contextuality of information. Information may be inescapably contextual. But that is not a problem. Using Lifiting rules one can decontextualise information (to one's own preferred context of course).

So that leaves the fact that there are no successful UFOs around, and yet it would be nice to have some good ones to play with. If they were around and easy to understand then we would have a context in which we would know it to be easy to merge information. I imagine database administrators would be really happy with this. No need to Refactor Databases anymore. We could start with a clean database layout using our UFO, and be assured that data would always fit nicely into it. Well life is life. It's not always as easy as we would like it to be. Humans have always been able to communicate, and this must have been possible without UFOs, since we are here.

So in the end the best way is to do exactly what the SemWeb does. Build a very flexible system on some very basic principles: URIs, triples, http, add an optional inferencing layer and allow the web to create a distributed framework. Grow UFOs! don't design them. [1]


  1. An I'll throw in a reference of my own Ruth Garrett Millikan's Language, Thought, and Other Biological Categories: New Foundations for Realism where she shows the fundamental biological nature of language, and how natural selection comes into play in grounding words in reality.

Very good stuff, I had the same feelings ;) If you had an rdf/semweb category feed for these kinds of articles rather than lumping them under "java" (!) you'd have something that'd fit on PlanetRDF...

Posted by Dave Beckett on April 06, 2006 at 11:35 PM CEST #

done :-)

Posted by Henry Story on April 07, 2006 at 06:13 AM CEST #

I recently discovered another reason people are mislead into believing in the necessity of UFOs: static programming languages such as Java, C++, etc... freeze the type hierarchy of a class. This is ok with programs that are compiled and written to run locally, as it can speed up and simplify the compilation process, and it makes for very fast execution. But it means that the designer has to get the type hierarchy right from the beginning. It is not easy to add a new class in between java.lang.String and java.lang.Object .

The semantic web on the other hand is build on an Open World Assumption, which means that it is built to allow for extensibility at every level. You can create the name of a relation, and only later define it. You can create the name of a class, define it in english, and only later define its supertypes. You can create a class, and have others define its supertypes, even. You can define what you need now, and when a UFO comes along, have your vocabulary tied into it.

You cannot of course guarantee consistency between all the statements made on the web, just as you cannot stop people from putting up garbage on their web site. But that is where natural selection comes in. Those pieces that work best together and are used to provide useful information will be used more often, and replicated more widely.

Posted by Henry Story on June 18, 2007 at 10:51 AM CEST #

Post a Comment:
Comments are closed for this entry.



« June 2016