LANG/GUIDE: Thought on compliance levels and WG goals (Sem/test please read too)

The following is a long message - but please bear with me - I think 
it important to a critical area of our WG.

As you might guess I've been thinking hard about the issue of levels 
of compliance, goals for the WG, etc. given the recent issues we've 
had with reaching consensus about these issues.
  I've come to the conclusion that there are essentially two main 
approaches to ontologies at the moment, and our group reflects 
members of both.  I think it is important to understand these 
different (not competing!) goals, and think about how we make sure 
our langauge is useful in both cases.  I am going to use a real world 
example, but please note that it is only meant to illustrate the 
issue, and not to imply either that that is the only place the issue 
comes up, or that the particular application is in any way more 
interesting than the others.

The set up:
  A large govt organization in the US uses a set of Description Logic 
tools on a daily basis to develop a monthly vocabulary that they 
release to the world.  The development process uses about seven 
full-time employees and some very complex (proprietary code).  The 
current vocabulary release is in XML, against a DTD also available.
  Why would they want OWL??

---------
1) Publishing the vocabulary:  the THESAURUS level
  What we've been calling OWL core or Level 1 is what I'm going to 
call the thesaurus level -- this is also our "RDFS on steroids"
  My research group at Maryland wrote a Perl program to turn their 
monthly vocabulary dump into DAML+OIL.   We used only the RDFS 
primitives, cardinality (they needed some restrictions to specific 
cardinalities - not just 1/0) and they needed unique property.
  So for publishing the thesaurus, they need a simply language that 
extends RDFS and looks like we've been calling OWL core -- they could 
put this into use tomorrow (and want to)

2) Building the vocabulary: the INFERENCE level
  But what about what we've been calling level 2?  Do they need that? 
The answer is yes.  They wouild like to find a way to share the 
ontology development with another organization or another group in 
their organizing a different tool kit - they are all using different 
tools.  So they would like a way to share their class definitions and 
etc. between tools.  For this they need a more extensive langauge.
  This is something they could not do right away, because their tool 
vendors would need to be involved - but if it was doable, they would 
indeed put pressure on their vendors to make this available and to 
make the tools interoperable.
------------

As a result, we see their are really two very different uses they 
(and many others) would like to put our language too.  One is to 
writing down the "output" of the ontology creation process (the 
thesaurus level) the other is to write down the inputs/specs of that 
creations (the inference level).

Many of the people I talked to at the WWW believe there is an 
immediate need for the thesaurus level, but when I discuss the 
inference level they see use there as well.  Some (for example the 
genomics folsk Ian works with) are already doing the inference level 
and others are seeing them talk about it and getting excited.

So my conclusion is that if our WG is going to fill "our client's 
needs" we must do both, we must discuss them clearly, and we must 
make it clear that the inference level builds on top of the thesaurus 
level (much as our thesaurus level extends RDFS in a functionality 
sense).

  I think the jury is still out on what the best way to describe this 
is (levels, document chapters, naming of functionality)- but I think 
it would be very valuable for us to start more clearly examining the 
real world issues like this with respect to how we'll make the 
language available - and how guide will write it up.

  As I'll say in another message later -- WWW has been amazingly 
exciting - a lot of real people want us to produce and will start 
using it immediately -- we will literally have millions of new OWL 
statements and thousands of pages using it soon after release -- so 
let's do it!!
  -JH





-- 
Professor James Hendler				  hendler@cs.umd.edu
Director, Semantic Web and Agent Technologies	  301-405-2696
Maryland Information and Network Dynamics Lab.	  301-405-6707 (Fax)
AV Williams Building, Univ of Maryland		  College Park, MD 20742
http://www.cs.umd.edu/users/hendler

Received on Saturday, 11 May 2002 20:49:19 UTC