W3C home > Mailing lists > Public > public-owl-dev@w3.org > April to June 2009

Re: UML-OWL Generator, A product to convert UML into OWL

From: Elisa Kendall <ekendall@sandsoft.com>
Date: Mon, 04 May 2009 12:24:29 -0700
Message-ID: <49FF40ED.4060407@sandsoft.com>
To: Bijan Parsia <bparsia@cs.manchester.ac.uk>
CC: public-owl-dev@w3.org
Hi Bijan,

Bijan Parsia wrote:
> On 4 May 2009, at 16:41, Elisa Kendall wrote:
>
>> Hi Bijan,
>>
>> I hadn't intended to point this out myself (since I'm assuming the 
>> folks who we've been exchanging email with have seen it and don't 
>> believe that it is an issue for their work), but thanks.
>
> It was the first think I pointed out, almost :) I'm not sure that 
> you'd really want to *thank* me, since I'm not clear what's 
> meritorious about your patent.
>
>> We had what we believed were some key insights years ago, confirmed 
>> with Grady Booch in fact, that led us to believe that in order to 
>> create a "proper" mapping from a UML model to OWL, you needed to 
>> understand more about the semantics of the model than might be 
>> available from traditional reverse engineering.
>
> I think that for any forward mapping from UML to OWL you can consider:
>     The generic semantics of UML (what most of the work on "reasoning 
> over UML class diagrams" focus on). I don't see how this is 
> legitimately patentable in the context of all the work that exists.
>     Trying to capture the "intent" of the diagrams (e.g., via some 
> analysis of the labels, using auxiliary ontologies, etc.) I guess I 
> could imagine *some* of these being patentable, if they were fairly 
> specific. I mean, "Translate to OWL then enhance" just seems too 
> broad. Obviously, if you can "enhance" in *some* way or another, 
> that'd be good. But I can imagine trying to enhance using NLP on the 
> labels and mapping them into wordnet, etc. or looking at code 
> instantiations of the model. But again, I'd want the procedure to be 
> pretty deterministic and specific.
The analysis requires more than this as it turns out, there are a number 
of patterns that tend to work at a high level, but many organizations 
have their own profiles, best practices, preferred patterns, and so 
forth that are also relevant, in addition to controlled vocabularies.  
UML is really very large, covering considerable ground on the behavioral 
side, and there are also an increasing number of metamodels and other 
profiles (e.g., SysML, SoaML, BPMN, etc.) that are relevant to any 
transformation, particularly if customers apply one or more of them to 
the model that they then want to transform to OWL.  We did originally 
use some NLP capabilities to extract terms, and may do so again, but 
even that has to be applied -after- you understand the patterns in the 
model.  Most other UML-OWL transformation approaches we've seen support 
logical models only -- the class diagrams, but there are several 
additional diagrams for which the semantics are quite useful (e.g., use 
case diagrams, state diagrams, component diagrams, etc.).

Note that we started working on these ideas in 1996, with an original 
target of UML-KIF, not OWL, which did not exist at the time.  Some of 
the basis for our work is derived from there, and included some support 
for rules originally, which we're also now revisiting in light of RIF, 
PRR at OMG, etc.
>
>>  This was early work to tease out some of the issues, including the 
>> need for not only a of the language metamodel but an ontology of 
>> critical terminology in order to "do the right thing".
>
> I guess I'm still not seeing what's special about the *technique* as 
> so described. (Admittedly, the description is pretty sketchy.) I could 
> see that the "ontology of critical terminology" might be valuable 
> (since, presumably, it'd make or break the translation), but that 
> seems to be something for copyright or a trade secret, not a matter of 
> patent. I mean, do you think your patent covers *any* use of an 
> auxiliary ontology in the translation?
Yes -- it's integral, in fact.
>
>>  We still use this approach in our tools, but have refined it 
>> significantly since 2000/2001 when we did the original research, as 
>> you might expect.
>
> Is there a readable account, e.g., a whitepaper? 
Nothing that would provide the level of detail you're interested in at 
this point - we haven't had the bandwidth to write one.  We're talking 
with JPL about doing so later this summer, though, if we have sufficient 
time and resources.  If we do, we'll certainly post it or submit it to 
an appropriate conference.
>
>> The approach covers the combination of the methodology and the 
>> transformation to OWL (or other things).  It predates ODM 
>> substantially, but our current work has been updated to support parts 
>> of the standard.
>
> I guess the question is whether one can use ODM without infringing on 
> your patent. Or perhaps what one must not do to avoid infringement.
There is nothing inherent in using ODM that would infringe on our 
patent.  Many UML tools support importing UML profiles and allowing 
users to apply them to their models, in fact.  It's only if one wants to 
import/export OWL, from a UML tool, using a metamodel, profile, and 
ontologies in the way that we've done that there is a possibility of 
infringement.
>
>> When we submitted our inputs to ODM (and since, with subsequent 
>> updates to the standard), we agreed to license any relevant patents 
>> to anyone who was interested at reasonable commercial rates.  That 
>> would include the one you found.
>
> Ok, so you selected "RAND" instead of "royalty free". If I wrote an 
> XSLT that translated UML diagrams into OWL that is aligned with a 
> foundational ontology (something along the lines of 
> <http://www.sfu.ca/~dgasevic/projects/UMLtoOWL/>) do I need a license?
No.
>
>>  We are also planning to contribute some of the work to an emerging 
>> Eclipse project, the Eclipse/MDT project, and hope to get the ODM 
>> metamodels, profiles, and APIs out in the Galileo release coming out 
>> next month, fyi.  None of those components require a license to our 
>> patent from a usage perspective.
>
> You mean that you've licensed to Eclipse the technology so they can 
> distribute it? But if someone released a similar project (e.g., for 
> NetBeans) they should come to you for a license?
We're donating it to the Eclipse foundation under the Eclipse license, 
and anyone can reuse it.  The pieces we are providing to Eclipse will be 
royalty free, and are useful without requiring application of our patent.
>
> Does TopQuandrent have a license? Does their UML conversion infringe?
They have not licensed our patent, but the last time I saw their tool, 
they rendered diagrams that were (1) not really UML, although they do 
use boxes and arrows, and (2) not editable.  Even if they produce UML 
(more likely XMI) which can be imported into a UML tool, they would 
likely not be infringing, but I'd have to understand what they are doing 
better to be sure.
>
> Do you think the UML-OWL Generator at least *prima facie* infringes? 
> If not, why not?  How about ICOM?
The UML-OWL Generator -could- be infringing, but there is not enough 
information available to be sure.  If it does, the patent examiner will 
probably find it long before it publishes.  The one we worked with was 
quite good, found a number of papers that helped him understand what we 
were doing, the potential benefits, precise wording of claims, etc.  I 
don't believe ICOM does (if it hasn't changed substantially in the last 
couple of years).

Best,

Elisa
>
> Cheers,
> Bijan.
>
>
Received on Monday, 4 May 2009 19:25:07 GMT

This archive was generated by hypermail 2.3.1 : Wednesday, 27 March 2013 09:32:57 GMT