[sc34wg3] New syntax for (binary) associations

Lars Marius Garshol larsga at garshol.priv.no
Fri Feb 1 17:30:32 EST 2008

* Dmitry
> I always  can define user friendly names for roles for  specific
> association types:
> tm: subject
>     - "Blog" @ blogging:has_post
> tm: object
>     - "Post"  @ blogging:has_post

This is straying perilously close to abuse of scope, I would say. Is  
"Blog" *really* another name for "subject"? I don't think so. And then  
when "Blog" also is object in another association type, things get  
really confusing.

There's also the issue that "subject" means something else in Topic  
Maps already.

> Some benefits:
> - simple predicate interpretation and ability to define  deduction
> rules with staying in the realm of associations (not like in tolog!)
> [...]
> o:has_parent(X,Y):-
>       o:has_mother(X,Y)  |  o:has_father(X,Y).

Your criticism of tolog is valid, but this is just a syntactical  
issue. One could have done this right in tolog by saying

o:has-parent(X : parent, Y : child) :-
         o:has_mother(X,Y)  |  o:has_father(X,Y).

> - simplified TMCL

Not really, unless you propose to make the standard roles required in  
all association types, which means existing ontologies will not fit  
into TMCL.

> - simplified TMQL


> - simple integration with CYC and other knowledge bases and inference
> engines

Because you don't have to declare role types? Maybe. Or maybe you  
could define the mapping with these two role types, or some similar  

> - interoperability with RDF (almost without any annotations)

We do this already in the Omnigator plug-in: the role types default  
to :subject and :object. You can do that when importing from RDF if  
you want. The only thing this *really* simplifies is TM->RDF conversion.

> Should I call this thing TM-Lite :)

I think you are really onto the core of the issue here. Yes, you  
probably should not confuse this with Topic Maps as we know them. :-)

--Lars M.

More information about the sc34wg3 mailing list