Re: Modal predication vs reification

Hi Peter,

TL;DR:  RDF is so weak that everything has to be shoehorned into resources and predications in a single interpretation.  Discussions that mention anything else need to be couched in a well-specified extension to RDF.  It is possible to shoehorn many predicates, including :believes, :is-told-that, :place-of, and :instrument into RDF as it currently stands.

True, but as you say:

Just because something can be shoehorned into RDF does not imply that it is a good idea to do so.

Indeed.
RDF is a language with interpretations in a model-theoretic semantics.
A model represents the resources (and their relationships) existing in the actual domain of interest.
Not every RDF graph can be interpreted in a meaningful way (wrt the actual domain of interest) by looking at its models.

If RDF-star is going to provide a meaning for embedded triples there are two options that I see.  The simple option is to make embedded triples be resources.  The complex option is to add a new kind of thing to the semantics of RDF.  I prefer the simple option, and a simple version of the simple option.

Agree.
There should be a total injective function from embedded triples to resources.

Anyone who wants to argue that a new kind of thing needs to be added to the semantics of RDF needs to come up with a firm account of how that new kind of thing works.  Any arguments about how to handle embedded triples that require a new kind of thing in the RDF semantics but that don't say how this new kind of thing works are impossible to analyze.
Arguments that the predicate :believes needs a "modal world" in RDF without saying how modal worlds work in RDF is in a similar quandary.  How am I to analyze whether this is the right way to go without such an account?

Agree.
That’s why I’m suggesting that it is a bad practice to have in a RDF graph triples meant to state “modal predications”.

My take on :believes, :is-told-that, :stated, :place-of, and :instrument in RDF is that they can all be regular RDF properties.  Indeed there is no other choice if they are to be modelled in RDF as it currently stands.

Agree.

Whether their objects are embedded triples depends on the predicate and the intended meaning of embedded triples.   Because of their quasi-uniqueness, embedded triples cannot represent statings and thus are not suitable as objects for :stated and maybe :is-told-that.

Agree.

Whether :place-of and :instrument take embedded triples as objects depends on some subtle representation decisions, including the relationship between an embedded triple and the relationship is corresponds to.  (One might want embedded triples to correspond to some sort of inherent truth, in which case they are suitable objects for relationships like :place-of; on the other hand one might want embedded triples to be separate from inherent truth, in which case there could be many :place-of relationships that one would want to represent so there needs to be an intermediate resource to keep them separated from each other.)

I guess that this would be the example of the triple:

:liz :spouse :richard .

which could be the embedded as object of a predicate :is-period-of :

:period-xxx :is-period-of < :liz :spouse :richard > .

but, since the embedded triple may correspond to two distinct events of marriage, this would be more appropriately represented as:

:period-1964–1974 :is-period-of < :liz :spouse-1 :richard > .
:period-1975–1976 :is-period-of < :liz :spouse-2 :richard > .
:spouse-1 rdfs:subPropertyOf :spouse .
:spouse-2 rdfs:subPropertyOf :spouse .

cheers
—e.