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

[Murray Altheim]

Quoting Simon Grant <asimong at btinternet.com>:
[...]
> For me, it is really very little to do with formality, and certainly
> nothing to do with formal logic (though I have nothing against formal
> logic). Where I come to Topic Maps from is from seeing that the TM
> approach is more *humanly* reasonable than, say, just RDF.
> (Not that I know anything significant at all about RDF.) I approach
> these matters partly from the perspective of human-computer
> interaction, and partly from cognitive science - how do we represent
> things in ways that are most closely aligned with human thinking, and
> therefore most easily understood by humans?

Hi Simon,

Coming in on what seems the tail end of this thread I both agree and
disagree with a lot of the comments. You must excuse Alex' dark
Norwegian-ness:  I don't think he's ever quite gotten used to all the
sunshine in Canberra. But both he and Kal are correct -- there is no
free lunch, you don't get anything for nothing, well, not quite
anyway. As Jack said, there are some advantages of Topic Maps over
other formalizations, particularly in the areas of subject identity
and subject-centric merging, and to that I'd also add that as an XML
model and syntax, it's IMO a lot easier to grasp and use than the
alternatives. RDF has a lot of very strange baggage both semantically
and syntactically, though you'll find a lot of religious adherents
if you join the church.

The basic TM model is very straightforward compared to say, frames, but
that means that the basic model is also missing some pieces that need
to be added to one's Topic Maps in order to provide some of the frame-
like functionality. I was able to construct large and complex ontologies,
but to do that I needed to create a framework for things like facets or
properties. Things like temporality aren't needed in many models and
aren't part of Topic Maps, so that needs to be created. Part of creating
these features is the process of formalizing a controlled vocabulary or
grammar, in other words, choosing a specific way of using the syntax to
represent very specific things... i.e., rules.

One of the strengths and weaknesses of Topic Maps as delivered in XTM
is that it doesn't already include these features (as there is no one
feature that doesn't have alternatives, even in "basic" logic). In
taxonomy there are many different models, some class-based, some
collection-based, some hybrids. There are many ways to represent time
and events, so in the end one just has to choose a method and stick to
it. You have to choose to be formal (i.e., consistent) if you want any
consistency in usage, which means you either educate and police your
users and/or design an interface to prohibit them from making mistakes,
(as if that were possible). It doesn't sound like either approach is
what you're looking for.

As to being "aligned" to human thinking, I think (without trying to
appear obtuse) that it's an unanswerable question. Whose "human
thinking" do you mean? (and pardon me, but I think "human" is redundant
here) And how would one determine any alignment? Epistemologically
it's impossible.

If you mean most closely associated with what some people call "common
sense" reasoning, the informal way that people generally view and
express ideas about reality (not implying here that there is one way),
then what you're leaning towards is in the area of my own research,
i.e., the cross between formal and informal modeling, where one might
build an application that has enough features of formal models to be
useful but not stray so far into formality (at least on the surface)
as to make it difficult for lay people to use. Again, here there is no
free lunch. It comes down to to educating your users and/or designing
an interface to minimize errors.

Starting back in the late 70s, Doug Lenat has been building the Cyc
ontology (see http://opencyc.org/), a common sense ontological system
advertising a "300,000+ term Cyc ontology, with over 1 million facts
relating the terms to each other", and in looking at that it's hard
to see how it could be used by one's mom, unless one's mom is an
ontological engineer. But coupling Cyc with a powerful linguistic
sentence analyzer it can do a decent job of interpreting natural
language expressions and generating answers to questions. But even
given its enormous complexity, it also makes mistakes. If your
application can deal with mistakes, great. I hope your application
isn't air traffic control, nuclear energy, or anything to do with
handling people's security, private information, or money. I'm
guessing you don't want to spend several decades building your model
though. Doug will tell you that Cyc is "Doug's ontology" and is not
meant to encompass all of human common sensical thought (much less
intuition), and he'd also certainly tell you there is no free lunch,
that you don't get the power of Cyc without that background ontology
and all that machinery. Cyc is now open though, so you could perhaps
build on top of it.

HCI doesn't address modeling directly, but it requires an underlying
model. If that model is weak, the interface is similarly weak. If you
choose to have no formal, consistent model, you'll have no consistent
interface, just confusion. Interfaces always act as constraints, but
few are built on explicit constraints designed as part of an
epistemologically-sound ontological model. Few UI designers ever have
to think about that kind of thing.

> So really, what I would be interested in knowing is, how many people
> around here are interested primarily in semantics and human
> comprehensibility of TMs, and would like (for instance) to explore the
> ways of using TMs to represent human knowledge in ways that *feel* most
> comfortable to the people involved. My guess is that the majority here
> are interested in the formalities, the logic, the proofs - maybe in
> something close to mathematical elegance. Again, I have nothing against
> that in itself. Just that, for pragmatic applications, for
> communication, it is really useful to have representations that are as
> intuitive as possible. That is a different kind of constraint to the
> set of formal and logical ones.

I have to agree with previous comments. It's an inappropriate
question; your presuppositions are wrong. This is a diverse
community and speaking with some experience of the matter, there
are as many different views here as there are people. Even given
that, I don't think that the majority are that interested in
formality and logic per se, they are more interesting in modeling
in an applications (rather than theoretical) setting. It's not a
group of mathematicians, though they are represented here too.

What you're calling "pragmatic applications" strikes me as a rather
strange notion, particularly as the term "pragmatic" rings with a
lot more meaning than you probably intended. What is a "pragmatic
application"? One that does something useful? One that performs
some sort of logic-based inferencing? Or one that does some sort
of intuitive magic, making machine inferences lacking any machine-
interpretable statements? The latter has been a broken promise of
AI for many decades, though I still see jokers trying to sell this
notion to the ignorant*. The "Semantic Web" holds that same vain
promise if you're willing to drink the Koolaid.

As to intuition, there is no one generalized human intuition, so there
can be no one generalized model. That idea died about 2500 years ago.
I'd be interested in hearing your ideas on intuitive constraints. That
sounds a bit like witchcraft to me. If can't be formalized or otherwise
expressed it must be a form of magic. Even if we ignore the magical
and mystical elements and stick strictly to the forms of expression,
we stray immediately from epistemology into the realm of language, of
linguistics. Very thorny, again, no free lunch. And my understanding
is that we're still decades away from computational linguistics
providing anywhere near the promises made by the AI types in the 1970s.

> Is there anywhere else where people of the kind I am talking about hang out?

No, sorry. You can choose a community to hang around with, but I've
never seen one that was of (relatively speaking) one like mind,
except religious cults. The closest you might find to the kind of
informality-without-baggage you seem to seek is probably the
tagging and folksonomy types such as at del.icio.us. But those
"systems" (and I use the term loosely) are simple classification
systems, not ontological models, and they raise a lot more problems
and issues than they actually solve, a free lunch perhaps, but
likely to upset your stomach if you look to them as nourishment
rather than simple entertainment...  like the McDonalds of KR.

Murray

* http://www.acmqueue.org/modules.php?name=Content&pa=showpage&pid=386
...........................................................................
Murray Altheim <murray06 at altheim.com>                              ===  = =
http://www.altheim.com/murray/                                     = =  ===
SGML Grease Monkey, Banjo Player, Wantanabe Zen Monk               = =  = =

       In the evening
       The rice leaves in the garden
       Rustle in the autumn wind
       That blows through my reed hut.  -- Minamoto no Tsunenobu