W3C home > Mailing lists > Public > public-semweb-lifesci@w3.org > January 2007

Re: [biont] Nice wikipedia page on ontology

From: Alan Ruttenberg <alanruttenberg@gmail.com>
Date: Wed, 24 Jan 2007 13:09:15 -0500
Message-Id: <C4C406E2-F328-4DC2-9726-17F0C8506B20@gmail.com>
Cc: public-semweb-lifesci hcls <public-semweb-lifesci@w3.org>
To: Phillip Lord <phillip.lord@newcastle.ac.uk>


On Jan 24, 2007, at 11:57 AM, Phillip Lord wrote:

> Phil> Hmmm. Sure I wrote more than that in my original email.

Didn't see more. But great that everybody jumped in :)

> Phil> Yeah, Robert has my main beef which is the distinction  
> between the
> representation language and the representation itself.

Yup. Though there is too often a confusion between the ontology and  
the representation. In some ways I think that it is unfortunate that  
OWL has "Ontology" in it's name. It doesn't really come with any kind  
of philosophical commitment, other than the OWL-DL POV that your  
ontology should be consistent and be able to do something more for  
you than being a parchment from which to read back what you wrote.

> Phil> The use of "algorithms" is clearly wrong and I don't think  
> that an upper ontology
> provides consistency checks, nor that an ontology needs one to be
> formal.

Algorithm's not a good word. Formal's not a good word. I'm not sure  
"Upper" is a good word (foundational seems better). But then, it's  
hard to find good words. I do think an upper ontology should state  
enough that a reasoner has a chance at making reasonable inferences  
that follow from the theory, including the flagging of  
inconsistencies (nonsense).

Actually when I viewed the article I hardly paid attention to the  
initial definition. I'm so used to there not being good definitions  
that I think I instinctively skip over stuff like this. What I liked  
was common terms section, the direction that the applications section  
is going, and the links to the foundational ontologies.

> Phil> Still, it's an early wikipedia entry. These things often  
> improve over time.

Less interested in the article, to be honest, than HCLSIG. I think  
we're still at a point where we don't all have enough common ground  
upon which to discuss ontologies clearly, and have been wishing we  
could get to a point where we do.

>>>>>> "Robert" == Robert Stevens <robert.stevens@manchester.ac.uk>  
>>>>>> writes:
>
>   Robert> 'd be inclined to agree with Phil. I don't where the bit
>   Robert> about "algorithms" has come from. The other mistake, I
>   Robert> think, is not to make the distinction between formality of
>   Robert> language for representaiton and the formality of the
>   Robert> ontology itself. The latter is, I think, a matter of the
>   Robert> distinctions made. One can make an ontology in a formal
>   Robert> language like owl, but still be informal in the ontological
>   Robert> distinctions made.

Yes. The distinctions being made by the ontology are what interest me  
at the moment. And if they are worthwhile distinctions, how to encode  
them in OWL. Background: I am trying to understand BFO enough to  
figure out what kinds of inferences should follow from statements  
made using it.

>   Robert> Formal ontological distinctions can be encapsulated in an
>   Robert> upper level, but upper level otnoogies are not necessarily
>   Robert> formal....

Quite.

My gut, at the moment, is that if the upper-level/foundational  
distinctions are well made, this will have a positive impact on the  
rest of ontology development. On the other hand, I'm finding that it  
is some work to be able to fully understand the distinctions and  
their implications (took me quite a while to grok OWL too). For  
instance I am puzzling over realizable_entity at the moment. Seems  
that, e.g.  disposition should be a defined class, something like "if  
there exists a process p in which x is a participant, and x is  
bfo:transformed during p, or something bfo:derives from x during p  
(during p means those things are also participants in p), then x has  
the disposition to be involved in p". But the OWL version doesn't yet  
say what relations dispositions can participate in. I've gathered  
that "inheres" is one, because realizable_entity is a dependent, but  
even that isn't yet stated.

BFO (in OWL) currently suffers from the fact that virtually no  
implications can be made. I'd like to remedy that. (Assuming I get to  
the point where I understand all the distinctions being made and feel  
they are reasonable - I'm getting closer).

>   Robert> Anyway, it is bad at almost any level

Do you not agree about what I see as the positives?

Happy to see you here!

-Alan
Received on Wednesday, 24 January 2007 18:09:33 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 17:20:22 UTC