TDL - Model theory take 1.

Hi Patrick,

I reply to some of your earlier questions below.

I attach a snapshot of where I am at with the Model Theory, along with some
comments on your text. The text I have produced so far is not either the
simpler or the more advanced text that I am planning to produce and hence is
not ready for insertion into the document yet. It's somewhere in between.
Once I have finished it at this level I will both simplify and embellish it
to produce the main text and the appendix text, respectively.

References for bit are:

The version of the RDF Model Theory I am working from is this one:

Patel-Schneider's document is:


> Does it help to split the interpretation process onto two
> "levels" or "phases" where first one interprets the idiom
> to obtain the TDL pairing and then, once the pairing is
> known, interpret the pairing to get the mapping.

Unfortunately not, see on.

> The latter phase seems (in my admitted ignorance) to be
> rather simple, since there is a one to one correspondence
> between a pairing and a mapping. And since the TDL pairing
> includes the type and lexical form, they are both present
> for the interpretation of that phase.

In practice we have to address the question of what happens when there are
two types specified. This motivates a significant part of the complexity. S
is significantly simpler in this respect, but buggy. (i.e. on the "as simple
as possible but no simpler" axis, S-B, IMO goes too far).

> As for the former phase, the interpretation of the idiom,
> does this really need to be grounded in the MT?

Yes, because the global type information can only be accessed through the
MT. e.g. a range constraint may be through any number of subProperty
relationships etc.

> Also, for idiom S/B, the mapping is not explicit, but is
> as implicit as the TDL pairing, as all that is identified
> in the graph is the lexical space of the datatype (with the
> datatype itself implied by the relation of lexical space
> to datatype) and the lexical form. How is TDL different?
> (asking this humbly so that I understand it myself)
> The node that represents the actual member of the value space
> is always constant, being the object node of the statement,
> which is either labeled with the literal or an anonymous
> node with rdf:value property defining the literal.
> > I think we bite the bullet and make noises about tails and dogs).
> Right.
> > I am currently planning to get writing on Wednesday. I might
> have some time
> > tomorrow.
> Fair enough. I plan to work on the final verbage today and tomorrow,
> adding in the discussion of the desiderada and some clarifications
> suggested by recent comments.
> I think I'll put together a separate summary of what I see as the
> shortcomings of S, not including it in the TDL proposal itself.
> Patrick
> --
> Patrick Stickler              Phone: +358 50 483 9453
> Senior Research Scientist     Fax:   +358 7180 35409
> Nokia Research Center         Email:

Received on Tuesday, 22 January 2002 05:29:47 UTC