Dear Keith,
first: congratulations for going that far with KIM and thanks for
your interest.
You are absolutely right about the behaviour of KIM Gazetteer. This was
designed as a features but in this case emanates as a bug.
I would not like to bother you with tech-details but still - here is a
brief description of the reasons:
The KIM Gazetteer is trying to "compress" the representation of the
classes and not in a very smart way, i have to admit. It just removes
the namespace part of the URI of the class, believing there are only
classes from the namespace of our World Knowledge Base.
Let me discuss this tomorrow with my colleagues and try to find an easy
fix. But, please let me know what is an appropriate time-frame for you
to get a fix of this issue.
We are quite under pressure because of an upcoming conference (ISWC06)
and our efforts until the end of the week are totally absorbed by this.
have a nice evening
borislav
Keith Suderman wrote:
> Hi,
>
> I've noticed strange problem with the way the KIM Client annotates
> certain entities. I am using the KIM Client in GATE, but I assume the
> problem is likely the same in the Populate Tool.
>
> Since we don't want to be dependent on PROTON, or any other upper
> level ontology, we use our own separate, standalone ontology to
> define our class hierarchy and then we use, what I think of as, an
> interface/mapping "layer" (for lack of a better term) to declare
> classes in our ontology as subclasses of the equivalent PROTON
> class. That is, our Nation class is a subclass of the PROTON Country
> class, our City class is a subclass of the PROTON City class
> etc. This way, if we needed to, we could remove the PROTON
> definitions and still have a valid ontology, or map onto a different
> ontology with a minimum of effort.
>
> This works perfectly, except for one small problem; if our class and
> the PROTON class have exactly the same name then KIM will annotate
> instances with the PROTON class. If the class names are different
> (i.e. Nation instead of Country, or CITY instead of City) KIM will
> use our class when annotating. I've run the Gazetteer only KIM
> Client and it seems to be the KIM Gazetteer that is annotating
> instances with the wrong class, although the instance references are correct.
>
> Is there something in the KIM Gazetteer that only matches the class
> name rather than the name and URI? Is there some way to get the KIM
> Gazetteer to always use our classes when annotating instances?
>
> One remedy I've thought of is to simply add a phase to the end of
> JAPE transducer (after final.jape) that matches instances that have
> been annotated with a PROTON class and changes the "class" attribute
> to our class name. However, before starting on this I wanted to ask
> if there was a better way, or if I am doing something wrong to start with.
>
> Cheers,
> Keith
>
>
> --------------------------------------------------
> Research Associate
> American National Corpus
> [EMAIL PROTECTED]
> http://americannationalcorpus.org
>
>
> _______________________________________________
> NOTE: Please REPLY TO ALL to ensure that your reply reaches all members of
> this mailing list.
>
> KIM-discussion mailing list
> [email protected]
> http://ontotext.com/mailman/listinfo/kim-discussion_ontotext.com
>
>
>
>
_______________________________________________
NOTE: Please REPLY TO ALL to ensure that your reply reaches all members of this
mailing list.
KIM-discussion mailing list
[email protected]
http://ontotext.com/mailman/listinfo/kim-discussion_ontotext.com