[topicmapmail] Not just Temporality in Topic Maps, but a whole approach

[Simon Grant]

At 01:16 2006-06-13, Alexander Johannesen wrote:
>On 6/13/06, Simon Grant <asimong at btinternet.com> wrote:
>>[..] I intended to ask
>>1. whether people shared the view that TMs are in some way closer to
>>human mental representation
>
>It's what drove me to TM in the first place, this fuzzy "it looks like
>a very human way of representing knowledge", but these days I keep
>telling people that it is easy to see that, but much harder to do
>that. :)

Well, indeed, I can see that point. In particular, I agree that one 
should beware of the enthusiasm of the convert, who may imagine that 
the new "thing" answers all his or her problems - maybe this is 
related to the "free lunch" syndrome. But when the (internal) hype 
has gone, when realism returns (as after marriage, perhaps :-) ) 
there is still the considered question "is this the best one for me 
(or my clients)?". And the answer can be, yes, for good reasons. But 
the analogy fails - there are good reasons to stay with a spouse 
through thick and thin, but if a better technology comes along - 
judged better in a considered, rational way - the main reason for 
fidelity is self-interest. All that investment of time, going to 
waste? That position as a recognised expert, forfeit? It can happen to us all.

>I think you struggle with the paradigm shift you get from coming from
>data modelling to ontological modelling or knowledge representation
>systems. In reality, they're not *that* different, but the subtle
>differences - especially where you can talk about the language in
>which you model - can be *very* persuasive to us; I can call this
>relationship by name! This means I can look for "all topics of type
>'fish', has a name like 'lung', occurres in 'south-america', and
>lives-in 'delta's " That's a powerful way to express "more humanly"
>something like (a poor ad hoc example) "SELECT * FROM animals, names,
>regions WHERE animals.type = 'fish' AND regions.id = 'south-america'
>AND animals.region-id = 'south-america' AND names.aid = animals.id AND
>animals.habitat = 'delta' SORT-BY names.name GROUP-BY regions.id"
>*phew* And that's only if the data modeller had any reasonable
>training in normalisation; it could be far worse.

OK, so what I am saying is, let's try to be considered and rational: 
neither on the one hand to think that some wonderful feature means 
that some new technology is flawless; nor to the rather cynical 
extreme (not held by anyone here) that it's all the same, that  there 
is nothing new or better under the sun.

>Topic Maps does lend itself to a more natural way of modelling things
>by simply being ontological based, but in all seriousness, modelling
>simple maps is easy. Once they grow, and especially once you want to
>actually *do* something with them, that's where it gets hairy. Don't
>let this early simplicity let you think it will *all* be easy. :)

Fair point. I don't. Another interesting comparison might be, how 
does the difficulty of learning the harder bits vary across 
technologies? Fine, no technology magically makes difficult things 
simple, but some technologies make things significantly easier or 
more difficult than others.

>Topic Maps in itself doesn't provide anything extraordinarie out of
>the box; it's a technology like many other, although one that *I* use
>because I can do things faster and more efficent in them.

Well, that's a good start, and I expect to agree heartily.

In fact, I'm not even arguing the case that TM is a "better" approach 
than alternative techologies. I'm wondering (perhaps not even 
asking), what ways of representation *within* TM are closer to the 
ways the client conceives of the domain. Because, no doubt at all, TM 
gives you enough rope to hang yourself with very effectively, so some 
ways are worse, no question. I say, perhaps not asking the question, 
because of course there may be no clear answer short of experience, 
and responsiveness to the client.

Simon