[topicmapmail] Contextualized Topic Maps.

[Kal Ahmed]

schulz@ivs.tu-berlin.de wrote:

>Tom P wrote:
>  
>
>>Yes, it is always a question what a "context" is and how one should be
>>represented.  But no matter what, a context must, it seems to me, have
>>to do with relationships between a topic and ....what?  Whatever it is
>>could presumably be represented by one or more topics or associations,
>>and the original topic would be related to the context via associations,
>>I would think.
>>    
>>
>
>In my opinion, a "context" ist more than just a scope, actually is
>knowledge, and may be represented by the same methods. In a tiny paper
>(unfortunately german only) I proposed context to be represented as n-ary
>associations, which somewhat can be compared to Lenats work on
>multi-dimensional representation of context.
>Why so? Because a (concrete) context relates a bunch of topics to define
>it. With TopicMaps you can define both, the meta-layer by typing and the
>definition layer by building "context"-typed n-ary associations. With
>reification you can use context-relations for scoping or associating with
>context-related topics.
>Ok, this may sound a bit complicated. But building a context as an
>association gives a stronger semantic to it. Related topics play specific
>roles in a context and by having explicit, context-typed associations, one
>could think of comparing contexts by the bound topics.
>On the other hand, it does not necessitate to introduce new techniques or
>elements to represent context in TopicMaps.
>  
>

I proposed something similar a while back: 
http://www.techquila.com/scope-proposal.html
Although it requires a lot of syntax (nothing in XTM is exactly compact 
;-), it does provide a mechanism for multiple dimensions of context and 
for simple combinations, and it only uses existing XTM syntax.

In practice, I have found that many applications can function very well 
with a simple approach to context that can be fully represented using 
the existing scope syntax (as long as what ever approach you take is 
consistent). However, recently I have come to realise that there are a 
number of very interesting applications out there that can only really 
benefit from a model of context that goes beyond the "toy".

Cheers,

Kal