Wednesday, May 18, 2005
concept of standardizing unique definitions for words
communication to http://groups.yahoo.com/group/jena-dev/
One of the responses to our note is:
Perhaps he means that it is not necessary
to use URIs to represent knowledge in an RDF-like way. E.g. I'm sure there could be a less complex
syntax for giving unique names to resources.
But the URI standard was there, and was leveraged for the purpose. Really, I consider URI problems to be just
silly spelling problems, which can be often times detected with lint-like tools
such as Chris's Eyeball. I think the
real difficulty is in understanding how OWL reasoning works, all the various
subsets of OWL, and the intricacies of why adding certain features causes
reasoning to become much more difficult.
These considerations are embodied in the API (e.g. all the different
rulesets and OntModelSpec's), which many people get tripped up on.
If
someone can come up with a knowledge representation which is easier to
understand and use, and does what most people want, then I think that's great.
I'm not knowledgeable enough to be able to say how feasible that is... I'm just an engineer, not a scientist ;) But personally, I think they've done an admirable job of taking a complex idea and implementing it in a flexible and extensible way (and writing lots of documentation and answering a bazillion questions on these mailing lists). As a developer, I don't expect APIs to be perfectly simple to use; I agree with Diego that Jena is most usefully thought of as a library for use by developers in creating tools which, one would hope, are easy to use (for their intended audience).
Reply…
Yes of course, the point I made was the knowledge representation does not "need" URIs. Certainly Schank's frames did not use URIs. KIF did not need URIs.
Sowa and Ballard's semantic primitives do not use URIs. In fact John Sowa has many time criticized TBL's RDF layer cake model of semantic web.
The concept of a URI is on the surface satisfying, but simply ignores the huge issues that come from natural science... see
http://www.bcngroup.org/area3/pprueitt/kmbook/Preface.htm
(It is easy to be dismissive of the natural science that I am referencing in this preface, but the point we make is that human cognition occurs as a physical and natural process and that the classicial logics have not been adequate.)
The problem with URI is not merely that many attempts to use them results in "misspelling". The problem is that the URI layers one more difficulty on the complete understanding of OWL that only a very few obtain.
The whole thing becomes a type of intellectual elitism were god like members of standards committees (who are not scientists but are engineers) create very difficult APIs that are constantly changing.
There are alternatives, like Intellidimension, where the data encoding is a simple four column database.... and the concept of RDF is represented without the Java wrapping that is imposed by HP, or some other group.
I found Intellidimension and included it in the BCNGroup Roadmap because of its high functionality and its compatability to the W3C basic standards.
So we use RDF, but we also use the Soviet cybernetic formalisms that are based on a more general triple < a, r, b > where a and b are simply "concepts", (or properties or attributes) and the r is a relational operator (of some type).