[topicmapmail] Reification and Type Instance

[Borislav Iordanov]

Hi,

Disclaimer: I'm really new to topic maps and I'm currently doing my first
model with them, but...

Besides the practical implications enumerated by Steve Pepper, I can't
really think of a reason why an association's type wouldn't be also the type
of its reifier. My only comment was that it begs conceptual clarification.
See below.

> But let's try to reason it through, and see where we wind up.
> 
> If I say:
> 
> employed-by(lmg : employee, bouvet : employer) ~ lmg-employment
> 
> then I'm really asserting that the 'lmg-employment' topic represents
> the same thing as the 'employed-by' association. That does kind of

You could also say that the lmg-employment topic represents the association
statement (as a topic map item) that represents the "employed-by(lmg,
bouvet)" association (that presumably exists independently of the topic
map). In that case the type of the association should *not* also be the type
of the statement (i.e. the information item as subject) that represents it. 

So, is the subject of the reifier the TM statement or the association (the
real world phenomenon) that the statement describes/refers to? 

One on hand the TMDM spec states (5.3.1):

<<
A statement is a claim or assertion about a subject (where the subject may
be a topic
map construct). Topic names, variant names, occurrences, and associations
are statements, whereas assignments of identifying locators to topics are
not considered statements.
>>

The first sentence implies that an association topic map construct can
become a subject, presumably via reification. Hence this implies that the
subject of the reifier is in fact the topic map construct. [The second
sentence I interpret as "Topic names, variant....and association *topic map
constructs* are statements..."]

But in the Reification section (5.3.4), I read:

<< For example, creating a topic that represents the relationship
represented by an association is
reification.>>

which means that the subject of the reifier is the relationship represented
by the association rather than the association construct itself.

So there seems to be some ambiguity lurking in there.

Deep metaphysics and epistemology aside, the distinction "statement
construct vs. phenomenon" might matter for the following reason: if the
subject of the reifying topic 'lmg-employment' is the association statement
rather than the actual association, it might participate in further
relationships differently. Because it would be a statement you would be
talking about, you could assert its degree of validity (true, false, a fuzzy
value in between), you could assert its logical dependence upon other
statements (i.e. pays-salary(lmg, bouver) => employed(lmg, bouvet)), all
things that make less sense if the subject is the actual relationship. 

> imply that they should have the same type. It's been argued that the
> statement-type relationship and the type-instance (for topics)
> relationship are different relationships, but I'm not sure that
> matters.

It might matter like this: if the subject of lmg-employment is the topic map
construct then its type will be the type of a statement/tm construct, it
will strictly have association constructs as instances and the information
attached to it might include syntactic constraints such as the expected
roles etc. Whereas the *distinct* type of the association (the phenomenon)
would simply represent the concept employment which may have other instances
that are syntactically different (in terms of roles involved etc.), and
which might also have occurrences as instances instead of associations.

Also, how is the association reifier identified - with a subject locator or
with a subject identifier? That also seems to pertain to construct vs.
phenomenon distinction.

Let me know if this makes any sense to you.

Best,
Boris