You raise some great points and distinctions. In my opinion, the application/system dictates the goals, objectives, requirements, ... which are to be accomplished (all of this was the subject of quite a few other emails over the last two days). Then, the ontology captures the relevant concepts, relations, individuals, rules, ... of the application/system. Lastly, you choose the language in which you express that ontology (and it does not only have to be one language, as long as you deal with interchanging data/results).
The language and its supporting infrastructure have to support the demands of the ontology and the application/system. For example, one consideration is the need for reasoning and inference which is available with RDF and OWL. Not all systems require this capability. Another consideration might be the need to do mathematical reasoning (x > y), where RDF and OWL clearly fail.
On your point of OWL being insufficient to express necessary engineering concepts, I wholeheartedly agree. It is a language with logic, but is not a commonsense knowledge base. There are many piece parts that currently exist (RDF/OWL/FOL, reasoners, SUMO or Cyc, WordNet, etc.). Each plays a valuable and different role in the ultimate solution. Any one of these, alone, is insufficient.
From: henson graves [mailto:henson.graves@xxxxxxxxxxx]
Sent: Sat 1/21/2012 2:52 AM
To: Westerinen, Andrea R.; 'Ontology Summit 2011 discussion'
Cc: 'Matthew West'
Subject: RE: [ontology-summit] Summit Engineering Tracks
Andrea,
Your note is exactly the kind of dialog I was hoping to get started.
It brings up what I believe is an important distinction for applied ontology. In this case to engineered systems and to engineering of systems (meta-engineering). The distinction is between the application, the language used to talk about the application, and the specific knowledge (ontology) represented within the language. OWL2 is a perfect place to discuss some issues. OWL2 is not only an ontology language, but a formal ontology language, and has the virtue that it has good reasoners. Starting from this about 4 years ago I started looking at and attempting to use OWL2 for representing product requirements and product designs. This starts in a series of papers with among others Ian Horrocks, and David Leal. It clearly is very promising. However, I have found that it is insufficient to represent the semantics needed for concepts such as part-whole relations. Certainly, one can introduce binary properties and call them part properties. But the language without extensions is unable to do a good job of representing a lot of engineering concepts. I am sure the ontologists would concur. So for me it is not a question of throwing OWL out, it is what semantic concepts are needed, how to express their semantics, and extensions to OWL are needed. Also, by the way, as an engineer I have been very much involved with using SysML to describe large scale systems and their interaction with the world. I view SysML as an ontology language, albeit one without a formal semantics. I also have been very much concerned with retrofitting SysML with a formal semantics and adding OWL class and role constructions. All to be able to build suitable ontologies for engineered products and more recently biomedical systems such as the human heart.
- Henson
From: Westerinen, Andrea R. [mailto:ANDREA.R.WESTERINEN@xxxxxxxx]
Sent: Friday, January 20, 2012 12:06 PM
To: Ontology Summit 2011 discussion; henson.graves@xxxxxxxxxxx
Cc: Matthew West
Subject: RE: [ontology-summit] Summit Engineering Tracks
I have recently written a paragraph summarizing "why (OWL) ontologies?" for a customer. It tries to address some of the points that Henson raises below.
Here it is with the identifying text removed:
Xxx requires the analysis, communication, comparison and [alignment] of [concepts] within and across authoritative tiers, addressing broad (high-level) to specific (low-level) enterprise environments. These requirements necessitate the creation of formal, semantically enabled models [of the concepts], and their identifying and supporting properties, relationships and individuals. Providing both a formal encoding and semantic richness allows normalization of the definitions (intent) and provides the ability to aggregate, compare, and reason over the [concepts]. These tasks and requirements are well-aligned to the goals and capabilities of an ontology-based approach. Ontologies, defined using W3C's OWL, can be created, stored and reasoned over using COTS tooling. In addition, many complementary supporting ontologies can be immediately imported, aligned and reused (such as the provenance and time/event ontologies from W3C, an ISO-3166 country ontology at downlode.org, and specific domain ontologies such as Xxx). Even if existing data is not in an ontological format, but is perhaps stored as a relational database, there is existing tooling to convert this data to an RDF encoding. (Taking this approach removes the need to use complex staging tables to mediate database information.) Using an OWL ontology as the basis for the Xxx Model allows the formalization of not only the core concepts (...) but also puts strong focus on the relationships between these concepts and the definition of formal-logic-based restrictions, facts/axioms, and rules. Using COTS OWL reasoners, logical analyses of the consistency, completeness and minimal set definitions are straightforward, along with the ability to align concepts and infer new data based on logical expressions and if/then (Horn) rules. Communication of a standardized, ontological, machine-understandable format [to environment-specific] translation agents will produce consistent, traceable and auditable definitions for specific end-point implementations.
From: henson graves [mailto:henson.graves@xxxxxxxxxxx]
Sent: Fri 1/20/2012 9:00 AM
To: 'Ontology Summit 2011 discussion'
Cc: 'Matthew West'
Subject: [ontology-summit] Summit Engineering Tracks
The track co-champions are soliciting input, participation, and references
for the two tracks on engineering of large systems and the resulting
engineered systems.
My interest for the engineering tracks is to establish dialog in the Summit
not only to identify engineering problems for which ontology can offer
solutions. But to go beyond that to dialog on what ontology results and
methodology can be applied and look at use cases for its application. Many
people in the industry of developing and using engineered systems are aware
that ontology may provide value to many of their problems and issues. But
the industry position is what ontology technology can help, how do we use
it, what are the benefits, and what will it cost. At least this has always
been the management response when I have taken proposals regarding ontology
forward in industry.
If this did not go to the general interest list, someone give me a pointer
to the right one.
- Henson Graves
|