[sc34wg3] Reification or Representation?
Lars Marius Garshol
Mon, 23 Aug 2004 20:36:20 +0200
* Martin Bryan
| As far as I can work out, Lars et al, in TMDM, restrict the meaning
| of the term reification to refer to the set of topics whose point of
| reference is a "Topic Map Construct" - some component of a topic map
| tree, represented in some concrete computing language.
| It would seem likely, though this point is not perfectly clear to
| me, that Lars et al believe that what such reified topics have as
| their subject is not some part of the topic map tree itself, but the
| nodes "identified" by that particular subset of the topic map tree.
Well, almost. The language may be a little confusing, but what's
intended is that if the topic reifies an employed-by association
between me and Ontopia the subject of the topic is my employment at
Ontopia (ie, the subject is *not* the association, it's the real-world
(yes, Bernard :) thing it represents).
| [My confusion with this is that I still don't see how I can make the
| subject of my topic the fact that an association has been set up
| between something with this role and something with that role
| ... but hopefully someone will explain that in terms I can
| comprehend one day.]
Actually, that's been discussed at length and the conclusion that was
reached in the end is that topic maps don't provide any machinery for
creating topics that represent other topics, either. So the answer is
that at present this is not explicitly supported. You can define a
convention for this yourself, and use TMQL to query it, but you really
will have to do stuff yourself.
| Steve Newcombe, on the other hand, seems to use reificiation in a
| much wider sense, whereby any topic defined within a topic map tree
| is a physical representation of some abstract concept.
Yes. (We, like you, are using "represent" for this.)
| It seems to me that the two uses of the term are not different, but
| that one is more restricted than the other in that it introduces a
| specific constraint that the subject being reified has already been
| recorded as a node in a topic map tree.
| Interestingly, the New Oxford Dictionary of English offers an
| alternative definition of the term reify - which might give us a
| small clue as to how we might make better use of the term: "make
| (something abstract) more concrete or real".
| The point that seems to have been missed is that a topic map is a physical
| (e.g. concrete) representation (if only in terms of bits) of an abstract
| concept. There can be many different physical representations of a given set
| of abstract concepts, depending on which language (HyTM, XTM, LTM, RDF, ...)
| is being used to represent it. We should be careful not to bias ourselves by
| using terms that are specific to a given language.
| (As an aside, I have returned from Ireland inside a cat, but not the
| one advertised in Dublin as the Seacat. We passed the Isle of Man,
| which reminds me that not all cats share common characteristics,
| such as tails. And when we refer to big cats we don't generally
| include sea-going cats in the same category. The point of this
| diatribe is to remind everyone that words mean what the context they
| are used in determine they mean, not something concrete - they are
| very impure abstractions for a concept.)
I'm with you on that, but on the other hand you explicitly used the
term "cat" here to trigger specific thoughts in people's minds,
because of your expectation of what people thought it meant.
I think we have a similar case with "represent" and "reify", except
that there are different groups of people who attach different
meanings to them. My view is that as we are operating within computer
science we should use the terminology of that field, right or wrong,
and define it as carefully as we need to in order to achieve what we
want. (A further problem being that we want different things, of
| Is there a solution to the fundamental dichotomy between the use of
| the term reificiation in ISO 13250 and TMDM?
There's several, which is the problem. :-)
| Is the fundamental difference in the use of the term that TMDM is
| trying to introduce a new concept something along the lines of "the
| purpose of reification within TMDM is to make concrete an abstract
| concept that has been introduced by relationships between the
| components of a particular topic map tree"?
Yes, I think it is. Note that the term has been in use in this sense
for decades already. RDF uses it to mean the same, as does XTM. (See
the last paragraph of 184.108.40.206 of XTM. This is basically about
untangling what the two first paragraphs are about from what the third
is talking about.)
| Am I right is suspecting that what is being reified within TMDM is,
| for example, the fact that a particular association has a given set
| of roles that are undertaken by a given set of subjects that have
| been represented by a particular set of topics?
Almost. What's being represented is the relationship, so that you
could take the topic reifying the association (the one above between
me and Ontopia) and attaching an occurrence of type "start-date" to it
with the value "2000-04-23" (or whenever the company was founded).
| In other words, is it the tree subtended from the node being
| referenced that is important, rather than the node itself, or the
| sum of the concepts identified by the nodes?
If I understand you correctly it is the thing represented by the
subtree that is important.
Lars Marius Garshol, Ontopian <URL: http://www.ontopia.net >
GSM: +47 98 21 55 50 <URL: http://www.garshol.priv.no >