ontolog-forum
[Top] [All Lists]

Re: [ontolog-forum] Interpreting OWL

To: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: Ed Barkmeyer <edbark@xxxxxxxx>
Date: Mon, 18 Oct 2010 12:25:29 -0400
Message-id: <4CBC74F9.50702@xxxxxxxx>
Sean Barker orignally wrote:    (01)

> The statement was made that anything that could be written in EXPRESS 
> could be written in OWL.
>    (02)

That statement is simply false.     (03)

But it is first important to distinguish between OWL2/DL and OWL2/full.  
OWL2/DL is an extended description logic language, which greatly limits 
what it can express.  OWL2/full is harder to characterize; it has 
essentially the full power of RDF, which allows you to express almost 
anything, but tells you nearly nothing about how to reason about it.     (04)

EXPRESS WHERE clauses (domain constraints and RULE constraints) do not 
in general have images in OWL, and in many cases you would have to 
invent the RDF terms in order to phrase them in RDF (and OWL/full).  
EXPRESS supertype clauses and SUBTYPE CONSTRAINT clauses, however, 
should be completely expressible in OWL.    (05)

John Sowa leaped in to tell us that you can't describe a cyclic graph as 
an OWL A-box population, because it will break the tableaux reasoning 
algorithm.  This is true.  You can, of course, describe a cyclic graph 
as an EXPRESS population, but not easily in an EXPRESS schema, and since 
EXPRESS has no associated reasoning algorithm, you can't make any 
statement about what an "EXPRESS reasoner" can do with it.  So if we are 
going to compare cheese with chalk, we will get no reasonable results.    (06)

> However, some of the constructs in EXPRESS, particularly those 
> concerning the cardinality and structure of relationships are not 
> directly obviously expressible in OWL, such as the distinction between 
> a bag and a set.
>    (07)

Is it clear in EXPRESS what the logical model of a BAG is?  What is 
described in the standard is an implementation model that is a 
sequence/list in which no significance is attached to the position of an 
element.  The alternative model is a mapping from the member-type of the 
BAG to the non-negative integers, but that would require the instances 
of the member-type to be well-defined.  Since EXPRESS is unclear about 
whether REAL numbers are or are not machine representations 
(floating-point numbers with limited mantissa length), and since EXPRESS 
is unclear about whether entities are compared by name (identifier) or 
by value (the content of selected properties), it is not clear how the 
mapping model of BAG can be applied to many domains.  I do agree that if 
you know what you mean by a BAG in EXPRESS, there is no OWL/DL model of 
it, and the same is true of LIST, but you can create your own model in 
RDF.     (08)

ARRAY is well-defined as a function from some subset of the integers to 
a domain.  That can be modeled in OWL, but you may not be able to do 
useful reasoning over it.  In actual fact, most ARRAY objects are 
actually computational representations of some mathematical construct, 
in which the operations are well-defined in something like Mathematica, 
but hopelessly out of the class of EXPRESS or OWL.    (09)

What I am saying is that these 'data structure' thingies in EXPRESS are 
essentially implementation models of some more abstract concepts, with 
the exception of LIST (sequence), which is a common natural structure 
that is logically annoying and not a native structure of computers.     (010)

> However, it should be possible to create a first order interpretation 
> of OWL such that an EXPRESS relationship is a subtype of 'thing', and 
> the relationship constraints are then OWL properties. EXPRESS Entity 
> and Type also become subtypes of 'thing'
>    (011)

Interesting.  According the EXPRESS LRM, a "relationship" is a derived 
concept of an "attribute" of an "entity".  The most obvious mapping of 
"attributes" of entities is then to OWL "properties":  datatypeProperty 
(for attributes with EXPRESS TYPEs) and objectProperty (for attributes 
with EXPRESS ENTITY types).    On the other hand, SC4 practice is to 
model most "relationships" with EXPRESS ENTITY (because of the SC4 model 
architecture, and because ENTITY was implicitly used to model SQL 
TABLE).    So, as David Price points out, this choice is based on a 
particular set of modeling conventions, not on the EXPRESS language 
concepts.    (012)

> This then allows one to construct a second order interpretation by 
> suptyping Entity, Relationship and Type as STEP generic entities, such 
> as Product, Version, View, Property, Property-Representation, 
> Representation-Presentation etc. That is, EXPRESS entities provide an 
> upper level ontology for STEP in OWL.
>    (013)

Yes.  The first attempt at this was ISO 10303-221, as I recall (thank 
you, Andreas van Renzen).  ISO 15926 is a much more interesting 
approach.  But feel free to try again.  Alternatively, you could look at 
making your would-be ontology a derivative of something that exists and 
is useful, like DOLCE.    (014)

> One can then create a third order interpretation, as is done in the 
> STEP Application Protocols, in which the STEP generic entities are 
> interpreted in the context of a business process, so that Product is 
> either a product (AP 203) a part (AP 214) or a technical data package 
> (AP 232).
>    (015)

Ah, yes, the STEP architecture.  Create a high-level entity called 
Product that has no required properties and a definition that doesn't 
distinguish a physical thing from the design for a physical thing, and 
then subtype it in each application to anything the application actually 
needs to model.    (016)

In short, if you start with a set of models made using the EXPRESS 
modeling language, and two or three vintages of SC4 modeling 
conventions, you can probably find some way to map them to OWL that 
might be of some use.  The principal values in this conversion are two:
 - the models will become accessible to a generation of formal modelers 
who will never want to learn EXPRESS (or the SC4 data modeling conventions)
 - the models and data can then be manipulated using RDF tooling, and 
thus become common fodder for 21st century academics, and some emerging 
industrial applications.    (017)

Note, however, that if these are your goals, DL reasoning with these 
models is not a critical concern, and perhaps not a concern at all, and 
preservation of the mathematical elements of the EXPRESS models, and 
many of the complex interrelationship constraints, will not be 
possible.   But then, the target RDF and OWL tooling can't do anything 
useful with them anyway.    (018)

Sean later wrote:    (019)

>     The point was rather that, if in the semantic web, one came across 
>something that said it was a AP214:Part which was a subtype of a STEP:Product 
>which is a subtype of EXPRESS:Entity, which is a subtype of OWL:thing, would 
>current approaches to the semantic web interpret it correctly?     (020)

Yes.  They all understand subsumption, and thus the AP214:Part will have 
all of the defined properties of the universal 'thing', as will 
AP214:Tool and NCI:melanoma.  If, however, you also define 
tool-is-used-in-producing-part as a subtype of STEP:Relationship and 
STEP:Relationship as a subtype EXPRESS:Entity, the Semantic Web engine 
will also it all the defined properties of 'thing', as opposed to the 
properties of objectProperty, and that may prevent any useful reasoning.    (021)

David Price provided a sensible response.  Basically, what he said is:  
what part of the EXPRESS model do you want to capture in OWL, and what 
do you intend to do with it?    (022)

All I would add is:  and is OWL useful for that purpose?    (023)

-Ed    (024)

P.S.  For the record, Uwe Kaufmann, Bernd Wenzel, David Price, and I 
developed a formal metamodel for the EXPRESS language as an OMG 
standard, which is finally about to be published in final form.  The 
approved final draft has been on the OMG server for a year at:
  http://www.omg.org/cgi-bin/doc?dtc/2009-11-05.pdf
The intent of this specification was to be a foundation for formal 
mappings from EXPRESS to OWL and UML (as the introduction says).  
Unfortunately, all of the participants in the original effort in 2006-7 
have moved on to other activities.    (025)


David Price wrote:
> Sean and others,
>
> First, the success in mapping from an EXPRESS schema to an OWL 
> ontology depends on the style in which the EXPRESS schema was written. 
> OWL itself has many uses and styles, so nothing new about that. The 
> reeper tool, of which I am the author, has an initial EXPRESS to OWL 
> mapping that is a 'structural mapping' largely aimed at enabling 
> exchange of data using RDF as a data encoding rather than Part 21 - 
> what some call a proxy ontology of the EXPRESS, rather than a rich 
> ontology of the domain of discourse. So, while I agree it has little 
> value on the Semantic Web to do not agree it's not a sensible thing to 
> do. I'm looking at a PLCS-based project where it enables a far 
> superior technical solution to data exchange because of its tight 
> integration with PLCS Reference Data which is modeled as OWL taxonomies.
>
> It's my contention is that there is no general EXPRESS to OWL mapping 
> possible if the requirement is to produce a semantically-useful 
> ontology. You must have knowledge of the meaning of the entity types 
> in the EXPRESS schema in order to produce useful OWL. Based on that 
> knowledge, a human would configure the mapping based on individual 
> entity types or common patterns in the schema and the tool could 
> produce useful OWL. So, my longer term aim for reeper was to create a 
> configurable mapping from EXPRESS to OWL to address the issues you and 
> others have raised. The supported pre-built configurations were to be 
> based on patterns of EXPRESS usage.
>
> Base to the SC4/STEP question - I imagine what whoever made the 
> statement meant was that for SC4 usage, there are very few cases where 
> OWL could not replace EXPRESS as the base modeling language for the 
> standards. That's certainly something I believe. In cases where OWL 
> doesn't do everything EXPRESS does, other approaches can supplement 
> the OWL (e.g. a lack of OWL support for cardinality can be handled by 
> an XML Schema derived from the OWL for the purpose of data exchange). 
> By the way, translating EXPRESS to UML first doesn't solve everything 
> - it turns out that UML doesn't support the full set of INVERSE 
> capabilities EXPRESS has.
>
> Cheers,
> David
>
> On 17/10/10 20:13, sean barker wrote:
>> John, 
>>
>>     Thank you for your answer on the relative expressive powers of OWL and 
>EXPRESS.
>>
>> Matthew, Ian
>>
>>     I was not actually proposing to write EXPRESS in OWL, although I believe 
>there is a tool by the name of Reeper that does the translation for you, but I 
>would agree it's not a sensible thing to do.
>>
>>     The point was rather that, if in the semantic web, one came across 
>something that said it was a AP214:Part which was a subtype of a STEP:Product 
>which is a subtype of EXPRESS:Entity, which is a subtype of OLW:thing, would 
>current approaches to the semantic web interpret it correctly? The reason for 
>asking this question is that, in my experience, let a computer programmer 
>loose with some constuct which is intended to have a conventional 
>interpretation and they will ignore the convention, and manipulated according 
>to its computational properties. My understanding is that "semantics" in 
>"semantic web" is concerned with the semantics of the logical operators, 
>rather than the semantics of the terms, such as instantiate from OWL:thing. 
>Hence, if one were to interpret OWL:thing as simply the modelling construct 
>EXPRESS:Entity, would subsequent subtyping to STEPProduct or STEP:Version make 
>further interpetation impossible? Or would this provide another form of Upper 
>Ontology?
>>
>> Sean Barker, Bristol
>>   
>>
>>
>>  
>> _________________________________________________________________
>> Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/  
>> Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/  
>> Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
>> Shared Files: http://ontolog.cim3.net/file/
>> Community Wiki: http://ontolog.cim3.net/wiki/ 
>> To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
>> To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx
>>  
>
>
> -- 
> Principal Consultant
> TopQuadrant, Inc.
> US Phone +1 336-283-0606
> UK Mobile +44 7788 561308    (026)

-- 
Edward J. Barkmeyer                        Email: edbark@xxxxxxxx
National Institute of Standards & Technology
Manufacturing Systems Integration Division
100 Bureau Drive, Stop 8263                Tel: +1 301-975-3528
Gaithersburg, MD 20899-8263                FAX: +1 301-975-4694    (027)

"The opinions expressed above do not reflect consensus of NIST, 
 and have not been reviewed by any Government authority."    (028)



_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/  
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/  
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/ 
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx    (029)

<Prev in Thread] Current Thread [Next in Thread>