[Sigia-l] Topic maps and IA

[Jonathan Broad]

On Mar 27, 2004, at 4:01 AM, Listera wrote:

>> With all due respect, OWL and topic maps are not the same kind of 
>> thing
>> as a relational database.  They're all data abstractions, true, but
>> they function at a fundamentally different *level* of abstraction.
>
> I'm not arguing that they are identical. In fact, my comments were less
> about technology than about IA's ability to reach out to other 
> professions,
> as they need to work with them on complex/corporate projects.
>

Forgive me, but perhaps herein lies our misunderstanding--Lars Marius 
Garshol isn't an IA reaching out, but a knowledge engineer reaching 
towards IA.  I know his work from my own interest in topic maps, as 
technology.  I thought your comment was true, but unresponsive to his 
essay, which is what I'd been commenting on.  Context is easily lost in 
email.

> Now, corporate/large-scale data often resides in an RDBMS. A category 
> of
> people attend to the creation, access and maintenance of that data. 
> This
> field is much older than digital IA. If you want to talk to them it 
> pays to
> a) know their language and b) be able to communicate with them through 
> not
> the latest alphabet soup of 'standard' of the month, but more 
> fundamental
> notions of abstraction they are all familiar with. One way to make 
> friends
> in a corporate room full of data/dev guys is not to start the 
> conversation
> with words like "ontology" and "faceted" even though the fundamentals 
> of
> those notions won't be foreign to them.

I agree completely; IAs are nothing if not translators, and translators 
have no business correcting the grammar or--heaven forbid--questioning 
the word choices of their interlocutors.  Of course, neither can we 
resist, sometimes...

But, as I noted a moment ago, *Mr. Garshol's article* is a case of a 
_data/dev guy_ reaching out to IAs--through the language of library 
science.  Which is a language many IAs are familiar with.

There is a serious irony here, Listera, which I can't help but note.  
Topic maps were created *precisely* to overcome the barrier that 
"different ways of saying the same thing" poses to communication.   So 
I think it sort of trumps the thrust of your critique.  It's also 
precisely the reason IAs should be interested in topic maps, and take 
the lead in educating other data/dev folk about the approach topic maps 
take to resolving the issue of terminological misunderstandings--using 
the appropriate terms for the audience, of course. :)  It's what topic 
maps are good at that no previous formal model has attempted to be good 
at.

I find the origin of topic maps, as technology, to be a rather charming 
story of terminological naiveté and creative misunderstanding.  Please 
forgive me if you've heard this story before, but it just seems too 
appropriate.

A bunch of data/dev folk (electronic publishing publishing experts, in 
this case) were working on a particular problem: the automation of book 
merging. Their system was working well.  Chapters, tables of contents, 
footnotes--everything was merging nicely.  Everything--except for the 
indexes.  What they found was that indexes to very similar books were 
very hard to merge.  It was almost as if, each indexer's "mental map" 
of the subject matter was subtly but decisively different...

At this point any librarian or indexer (or user-centered designer) 
would've been shaking their head at the perplexed look on the computer 
scientists' faces.  It's well known in some communities of practice 
that two expert indexers, presented with the same, very short article 
to index, will tend to produce almost *completely different* indexes.  
This doesn't invalidate the practice of indexing, of course, but 
clearly it poses a problem for the automatic merging of them.

Where librarians would've laughed knowingly and moved on, however, the 
computer scientists refused to be put off by the misprision embedded in 
different indexer's views of the world.  In this case, I think their 
stubborn refusal to simply give up on the irreducible complexity of 
human discourse was a stroke of genius.  Librarians are far too 
accustomed to doing things the hard way.  (This can also be a 
strength.)

So anyway, I find topic map terminology worth learning, because its 
whole purpose is to bridge terminological differences.  The formalism 
of topic maps are unique, although they draw productively on the 
histories of both computer science, librarianship, and AI.  In the 
particulars, it's nothing too original.  But it is very close to 
solving the *right problem*, where other approaches fall short--from my 
perspective as an IA.

I thought Mr. Garshol's paper did a fine job of *showing IAs* why topic 
maps are important.  How IAs can convince *DBAs* of the benefits of 
topic maps is another matter.

>
>> The terminology you object to isn't new, it's probably the longest 
>> part of the
>> history of artificial intelligence research.
>
> My point exactly; one way to lose friends in that corporate room full 
> of
> data/dev guys is, in fact, to bring up the subject of artificial
> intelligence. :-)
>

Yeah, but topic maps actually work. ;-)

Jonathan