Re: OWL "Sydney Syntax", structured english

On Thu, 2006-11-30 at 19:07 +0100, Kaarel Kaljurand wrote:
[...]
> just to clarify a little:
> 
> what I think I've managed to prove with the OWL verbalizer
> (http://www.ifi.unizh.ch/attempto/documentation/OWL_to_ACE/) is that
> a Description Logic like SROIQ can be verbalized in acceptable and
> understandable English given that the input ontology uses English transitive
> verbs for the property names. Note that I'm only talking about object properties
> and not data-valued properties.
> 
> I.e. if one wants to discuss his/her ontology with somebody how doesn't
> understand any of the proposed OWL syntaxes (RDF-based or not) then
> the ontology should first be fixed to have verbs as object property names,
> and then verbalized.
>
> Without this fix, the verbalization would not be understandable. (A compromise
> is to use "verbs" in the form "has+Noun", the verbalization would not suffer
> that much.)
> 
> I just hope that the benefit of having a nice English verbalization outweighs
> the burden of having to apply this fix.

Changing the URIs in an ontology is about the most burdensome thing
you can ask an ontology developer to do.

It's *much* less expensive to add transitive verb labels.

Your earlier reply to this suggestion was:

> The idea of our verbalization is to produce an English text that can be
> parsed back into the official OWL representation without any loss in meaning
> with regard to the original OWL file. Using labels would break this
> design decision
> immediately.

But now you're saying that your verbalization only partially understands
OWL, since it doesn't handle datatype properties (and it only handles
OWL DL, not all of OWL). So the round-trip requirement is only going
to be met in a very constrained set of cases.

I think your tool will be much more useful if it exploits transitive
verb labels.

In particular, is useless to me if as long as it requires me to change
the URIs in ontologies; especially since, in many cases, the ontology
I am interested in is not where I have any ability to change its
URIs. But I can always add labels (assuming your tool isn't constrained
to one source of data about the ontology in question).


-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/
D3C2 887B 0F92 6005 C541  0875 0F91 96DE 6E52 C29E

Received on Thursday, 30 November 2006 18:19:09 UTC