FW: Reification thing questions

--
               
Patrick Stickler              Phone: +358 50 483 9453
Senior Research Scientist     Fax:   +358 7180 35409
Nokia Research Center         Email: patrick.stickler@nokia.com


------ Forwarded Message
From: Patrick Stickler <patrick.stickler@nokia.com>
Date: Tue, 05 Feb 2002 10:46:46 +0200
To: Seth Russell <seth@robustai.net>, Pat Hayes <phayes@ai.uwf.edu>
Cc: RDF Logic <www-rdf-logic@w3.org>
Subject: Re: Reification thing questions

On 2002-02-05 7:14, "ext Seth Russell" <seth@robustai.net> wrote:

> From: Pat Hayes (phayes@ai.uwf.edu)
> see: http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2002Feb/0107.html
> 
> Let me take a shot at these:
> 
>> Further to my last message, let me suggest some key questions to ask
>> about the things that a reification is supposed to describe, ie what
>> _:x is intended to denote when we write
> 
>>   _:x  <rdf:type>      <rdf:Statement> .
>>   _:x <rdf:subject>   <subject> .
>>   _:x <rdf:predicate> <predicate> .
>>  _:x  <rdf:object>    <object> .
> 
> It could denote an actual appearance of the triple in some document, at some
> time, authorship, etc ...
> But the actual node you mention above says nothing about an appearance, so
> the example is kind of skew.  A better example of an appearance of a triple
> would include at least one other predicate about the appearance.

This is the way I am using/planning to use RDF reification.

Each bNode of rdf:type rdf:Statement denotes a "statement-in-context"
(where the context may be null) where one may specify the
authority, source, time, possibly some scopes, and even an authentication
certificate.

I also see this as a way to address assertion versus attribution
(i.e. Bob asserts the moon is green vs. Stan asserts that Bob
says the moon is green, etc. by utilizing an ontology that makes
such distinctions explicit.

Given this contextual information about statements, one can define,
for a given query, a form of pre-query that selects statements based
on context -- such as all statements from trusted authorities which
are assertions (not just attributions) and which were made before
a given date and are valid for a select set of scopes; and only
for those statements do I wish to make a given query.
 
>> 1.  Does it make sense to say that _:x has properties like date it
>> was asserted, who wrote it, where it was imported from?
> 
> Yes.

Absolutely.

>> 2.  If there is only one of these things, where is it? Does that
>> question make sense?
> 
> If the thing to which you refer is the thing denoted by the node identified
> with _:x above, then it must needs occur in some RDF document.  But there
> are probably many of those occurrences, not just one. So if there must be
> only one of them, then maybe it would be the ideal triple in the sky .. and
> your answer is in Plato's Heaven .... but I hope that is not the way it will
> go down.

I think this question is similar to are ("5", xsd:integer),
("05", xsd:integer), ("000005", xsd:integer) or even
("00101", foo:binaryInteger) all the same thing -- namely
the integer 5.

Insofar as all three TDL parings denote the same value, they are,
but they are not the same thing in the graph. They are different
expressions of the same thing.

Likewise, one may have multiple rdf:Statement typed bNodes representing
multiple reifications of the same statement -- possibly with different
context attributes/properties -- and yes, they all are referring to
the same expression of knowledge, the same statment, but also describing
the context or attributes of that expression as well.

If you couldn't have multiple reifications of the same statement, how
could different people say different things about the same statement?

E.g., both you and Seth and I may all independently assert that
the sky is blue. The statement itself doesn't say who asserted it.
It's just a triple. We need three separate reifications of the
statement/triple to provide the context of authority. And we
can't be forced to make those authority statements within a single
reification because this knowledge may in fact be used by a fourth
party who is syndicating all three sources without any of us
being aware that our assertional and reified statements are being
used together in the same knowledge base, thus each of us has
reified our own assertions independently and thus we have three
reifications.

Besides, just how would you constrain an RDF graph to have only
one reification per statement?

>> If not, what kinds of question would make sense applied to the thing?
> 
> ... who said it, in what document, how much one trusts it, what might entail
> it, what it might entail, is it true, is it false, is it a lie, do you like
> it ... stuff like that.

also

When was it said?

What is its scope of relevance/visibility? (language,
product, region, privacy level)

What is its authentication certificate?

What is its period of validity (when does it expire)?

What is its priority/strength/ranking?

Should it override other statements with same S and P but
different O and which have lower priority/strength/ranking?

Is it explicitly asserted (by a human) or inferred? (i.e.
can it be discarded without loss of knowledge, even if
the knowledge is implicit in one or more axioms?)

Is it transient? (in contexts for run-time syndication of
knowledge from e.g. mobile clients for the purpose of
service/resource discovery, knowledge that should be purged
from the knowledge base afterwards)

...

Need I go on? ;-)

>> 3. Can the thing be located in an RDF graph?
> 
> Both the thing denoted (the appearance of a triple) and the thing denoting
> (the node you quoted above) are located in RDF graphs .. though not always
> the same graph at the same time.

I agree.

>> If so, and if that graph is in a document (with a URI),
>> is the  thing in the document?
> 
> Depends: one document could be describing the occurrence of a triple in
> another document or in the same document.

And in fact, that would be expected for reifications describing
statements made by some third party.

>> (Could it be accessed via the URI?)
> 
> I don't know how to do that.

If you mean that the reified statement bNode has associated with
it a property such as rdfx:source that has a URI as its value,
then the answer to this question is, yes, but only if the URI
is resolvable. One would expect that it is, but one never knows
in this "contemporary" era ;-)

>> If the document is copied, is the thing in the graph also copied?
> 
> Yes.

Hmmmm.... which document. If the reification is in the document,
then yes, the reification is copied if the document is copied.
If the reification is in some other document. Then of course it
wouldn't be copied.

>> If so, is the copy of it [some] other thing or is it the same thing?
> 
> It's some other thing.

Right. A copy of a reification is another (redundant) reification.
 
>> If another, does it also satisfy the description  in the reification above?
> 
> Depends on how precisely the description is qualified.
> 
> For example:
> 
> foo:bar goo:dar poo:sar.
> 
> [
> rdf:type rdf:Statement;
> rdf:subject foo:bar;
> rdf:predicate goo:gar;
> rdf:object: poo:sar;
> ex:time "9:15PM"
> email::mid  0$657ba8c0@c1457248a.sttls1.wa.home.com ;
> ex:documentLocation  :SethsOutbox
> ]
> 
> The description above describes the triple as it existed momentarily in my
> out box.  It does not describe the copy of that same triple as it exists in
> your inbox.

Uhhh... now I'm confused as to which "thing" we are talking about.
I thought the "thing" was the bNode with rdf:type rdf:Statement.
You seem to now be equating "thing" with the triple. Or have I
just gotten gonzo confused ;-)

If the "thing" is the reification, and if the reification is copied,
then of course the copy describes the original statement as accurately
and completely as the original reification. Why wouldn't it?

Cheers,

Patrick

--
               
Patrick Stickler              Phone: +358 50 483 9453
Senior Research Scientist     Fax:   +358 7180 35409
Nokia Research Center         Email: patrick.stickler@nokia.com


------ End of Forwarded Message

Received on Tuesday, 5 February 2002 06:43:50 UTC