Re: Typing: was Re: DBpedia as possible project

Hi Lars,

On 4 Nov 2010, at 4:26 PM, Lars Heuer wrote:

[snip]

> And it's also okay to use type-instance relationships to indicate the
> profession of a person (personA isa astronaut). That this should be
> better modelled as an association where a person plays the role
> "astronaut" is a bit nitpicking. If it fits, someone may define:
> 
>   bob-the-builder isa construction-worker.

Actually, I was saying that you might actually have both statements in the same map, using whichever was most expressive for whatever you were trying to say. ;)

Not clear on what the TMDM position of this sort of thing might be, but it certainly fits the way things conceptually work in my implementation.  It's just another way of talking about how things are related.  Come to think of it, it might be worth changing one of the visualizations I use to be biased towards displaying properties of the role type if it's defined rather than just defaulting to the type of the proxy instance that's the start of the link.

All this typing was worthwhile after all! :)

> No problem at all. But could be a problem if "construction worker" is
> defined as
> 
>   construction-worker ako person.

Because construction-work is a narrower specialization?  What if they were reversed?

> Anyway, as others have stated, there is no absolute truth and each
> topic map is context-dependent/domain-specific, so the "Bob the
> Builder" statement is valid in several contexts and false in some
> other contexts.
> 
> Using type-instance relationships more carefully may generate less
> false positives. And using type-instance relationships for facts which
> don't change (or unlikely change) may make topic maps more robust
> (i.e. time independent) or gain a bigger audience which agrees with
> the statements made in the topic map.


Thanks for your thoughts and participation in the long-winded discussion.  Much appreciated!

Cheers,

ast
--
Andrew S. Townley <[email protected]>
http://atownley.org