ontolog-forum
[Top] [All Lists]

Re: [ontolog-forum] Semantic Web shortcomings [was Re: ANN: GoodRelation

To: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: Ron Wheeler <rwheeler@xxxxxxxxxxxxxxxxxxxxx>
Date: Sat, 16 Aug 2008 11:20:42 -0400
Message-id: <48A6F04A.5000803@xxxxxxxxxxxxxxxxxxxxx>
Very clear and to the point response.    (01)

I am not sure that structured English is absolutely required.
A GUI that allows the SME to create a new instance of an object and link 
that to other instances of objects as well as specify the properties of 
the instance and links to external references is probably all the SME 
needs for a basic system.    (02)

If this allows searching and navigating through the ontology to select 
the objects and instances, in a convenient way, that should meet most of 
the needs.    (03)

In general, I think that the academic side should have only limited 
responsibility for the filling up of the ontologies. This is the role of 
the IBMs, Suns, General Motors, Pentagon, Artifact Softwares, etc. Once 
the research community decides on a good architecture to represent the 
reality of a situation or set of entities, it is up to those who have a 
business or economic reason care about this should pay to have the 
details added.    (04)

What they decide to contribute back to the overall community will be 
based on their need to share the cost of maintenance and their 
understanding of the value of having their particular view of a part of 
the universe adopted as a "standard" weighed against the loss of a 
proprietary asset.
This is the basis of the open source movement. Some academic and vanity 
products get created but the big open source packages have corporate 
support when the corporations best interest is served by sharing.
Where it is not, the extensions to open source ontologies will be 
retained as proprietary to the creator and they will get any benefit and 
have to pay all the costs.    (05)

In my case, if I come up with an ontology for the process industries, I 
will have to decide if it serves our best interests to release this and 
hope that the big companies adopt it (and pay us to extend it or apply 
it) or to keep it private and resell it to clients one at a time. Not a 
simple decision to make but not hard to understand.    (06)


Knowledge engineering work  is very much like instructional design work. 
Someone gives you a stack of manuals, drawing, equipment specs and 
manuals and provides you with access to the engineers, maintenance 
supervisors and operations supervisors. You take these and figure out 
how to teach someone all about the process and what they need to know to 
do their job safely and effectively.
Coming up with a framework and basic principles that drive the ontology 
will require input from academic sources (here, books, articles, paid 
research, etc) as well the traditional resources listed above.    (07)

I also do not expect to get all of it right the first time.    (08)



I have been a consultant long enough to appreciate the joke: "A 
consultant is someone who borrows you watch, tells you what time it is 
and keeps your watch."    (09)

It is very infrequent that you get to tell a client something of great 
importance, that they did not already know.
Often, your greatest value is in helping them see an issue in the right 
context.
If you are good at consulting, you often just give them someone who 
cares about their problem(for $$$) and gives them someone at their level 
to discuss the issue until the choices and action items become clear.    (010)

We certainly do not pretend to know anything about the chemistry or 
physics or engineering of process plants. We are very good at organizing 
information and presenting it back to the client in a way that they can 
judge whether we understand it or not. We also have some experience 
which helps put their priorities and operation issues in perspective.    (011)

Ron    (012)


Ed Barkmeyer wrote:
> Ron Wheeler wrote:
>
>> I am not sure that ontology can be applied to high value projects 
>> without some elements of AI.
>
> I always assumed that "axiomatic ontology" was one branch of AI.
>
>> I am looking for intelligent agents (that are not humans - which I 
>> think, implies artificial) that can draw actionable conclusions in 
>> complex situations faster than humans.
>
> I think that is the general idea in knowledge engineering.  And it 
> also applies to the Semantic Web.  The idea is to allow the engine to 
> discard the chaff in the Google results and present you with only the 
> grain.
>
>> The other side of the equation is the tool sets that make it easy for 
>> a SME (Subject Matter Expert) to transfer his or her understanding of 
>> the universe into a model.
>
> There is work going on in that area, but it is definitely not ready 
> for prime time.  There are 5 or 6 "structured English" projects, but 
> they don't really help the domain expert become a modeler, which is a 
> skill.  And there is the automated text analysis work, which keeps 
> getting better and better, but still isn't good.  (Natural language is 
> just too messy.  It is amazing that we ever understand one another.)  
> The hybrid approach starts with a "sketch ontology" of the domain, and 
> then uses either the expert or the text to refine it.
>
>> I failed to realize that the gaining of PhDs was the only goal in 
>> life. (Too much time spent in the real world!)
>
> That is really a Catch-22.  Much of the government grant money goes to 
> academic institutions.  Most of the research, and almost all of the 
> product, of academic institutions is done with student labor.  Capable 
> students are willing to perform the labors of Hercules for a pittance, 
> as long as it leads to a degree in a few years.  Degrees are granted 
> for advancing the knowledge base of the science.  If a reputable 
> institution takes the money to build an ontology, it can't find a 
> competent student to do the work, because it can't give a degree for 
> recording existing knowledge.  So the ontology work has to be coupled 
> with some ostensible technological advance, and the latter becomes the 
> dominant theme of the work (and usually vitiates any value there may 
> have been in the ontology).  The only way the work actually gets done 
> is under some other umbrella: the student is given the task of 
> building the Augean stables ontology, "as a learning experience", 
> before he will be allowed to do the exciting work that will lead to 
> his advanced degree.
>
>> The SME has the job of building the model of a fair amount of the 
>> universe.
>> This SME might only have a high school education or junior college 
>> degree but he is the guy most knowledgeable about the relationship 
>> between valve 295 and the rest of the universe.
>> His knowledge may save a petrochemical plant as much money in one day 
>> as a PhD makes in a lifetime and may save many more lives.
>
> And all that is required is a skilled knowledge engineer (with or 
> without Ph.D.) who gets to talk to the SME, can find out that valve 
> 295 is significant, can determine from that conversation how it is 
> related to the rest of the plant universe, and can properly add that 
> knowledge to the ontology.
>
> I am reminded of the senior manager who asked why no one in his 
> organization had made the model we presented.  Voice from the back of 
> the room:  "Because it's nobody's <expletive> job to know how, and 
> nobody's <expletive> job to do it."
>
> There is always a reason why certain things don't happen (and another 
> well-known substance does).
>
> -Ed
>    (013)


_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/  
Subscribe/Config: 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 Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx    (014)

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