Re: terminating 'imports' loops (messages/events)

From: Dan Connolly ([email protected])
Date: 03/23/01


pat hayes wrote:
[...]
> I also want to formalize web protocols, but I'm going to try to do it
> by creating an ontology of web events first, rather than go looking
> for a new logic.

Yes... events/messages... exactly.

[...]
> Now, that "true"
> back there is good oldfashioned logical truth in an interpretation,
> so what we need here isnt a new logic, but a way to ground assertions
> in the intended world of web actions. Its grounding, not reification,
> that we need here.

OK, I'm happy to talk about it that way; that's what I meant.

> Plus of course some language to do the
> action-describing in, which I suspect will be pretty minimal; I bet
> we could do it as a DAML+OIL ontology, in fact.

That's what I was trying to do. Let me try again
in a bit more detail...


(rdfs:Class sw:Message) ; the set/class of message-events in
			; the semantic web

(rdfs:subClassOf http:Request sw:Message) ; HTTP requests and reponses
(rdfs:subClassOf http:Response sw:Message) ; and responses

(rdfs:subClassOf http:OK200 http:Response) ; HTTP responses with status
				; code 200 "OK, here's the content..."


(daml:UniqueProperty http:body) ; the bytes carried in the body
				; of such a message
(rdfs:domain http:body http:OK200)
(rdfs:range http:body OctetSequence)

(daml:UniqueProperty xml:parse) ; the (root element of the) XML document
			; denoted by some bytes
			;(I'm simplifying over some character encoding
			; and MIME type issues for this discussion)
(rdfs:domain xml:parse OctetSequence)
(rdfs:range xml:parse xml:Element)

(daml:UniqueProperty rdf:parse) ; the set/class of RDF
				; statements/triples/formulas/sentences
                                ; denoted by some XML document.
(rdfs:domain rdf:parse XMLElement)
(rdfs:range rdf:parse rdf:Class)
(=> (rdf:parse ?xml ?statements)
    (rdfs:subClassOf ?statements rdf:Statement) )


; so... an HTTP 200 OK message ?m says a statement ?st if
;  ?st is in the class of statements denoted
;  by the the xml document denoted by the body
;  of the message.
(<= (sw:says ?m ?st)
    (and (rdf:type ?m http:OK200)
         (rdf:type ?st (rdf:parse (xml:parse (http:body ?m)))) ) )

; now back to daml:imports
; if a DAML ontology is an HTTP resource, think of
; it as a set of HTTP messages from that resource.
; (you can think of an HTTP resource as an agent, in this sense).
(rdf:Property sw:from)
(rdfs:domain sw:from sw:Message)

; I guess the axiom for daml:imports that I gave
; in my previous message was oversimplified.
; I guess it should really work like this...

(rdf:Property sw:entails)
(rdfs:domain sw:entails rdf:Class) ; a set/class of messages entails...
(rdfs:range sw:entails rdf:Statement) ; a statement

(forall (?m1 ?m2 ?ont1 ?ont2 ?f)
        (=>
          (and (sw:from ?m1 ?ont1)
               (sw:from ?m2 ?ont2)
               (sw:entails (setof ?m1) ?f)
               (sw:entails (setof ?m2) ^(daml:imports ,?ont2 ,?ont1))
          )
          (sw:entails (setof ?m1 ?m2) ?f)
        ) )

; where entails and says are related by...

  (forall (?msg ?msgs ?f1 ?f2)
    (=> (and (rdf:type ?msg ?msgs)
             (sw:says ?msg ?f1)
             (wtr ^(=> ,?f1 ,?f2)) )
        (sw:entails ?msgs ?f2) ) )

[...]
> I'd be very interested in any other ideas you have in this area, by the way.

I'm trying to keep my thoughts on this near
  http://www.w3.org/DesignIssues/Toolbox#Assertion

The HTTP stuff above is basically my larch trait
  http://www.w3.org/XML/9711theory/HTTP.lsl
re-written in KIF/daml/rdf, and the says/entails
stuff is what I wrote in
  http://www.w3.org/2000/07/document-maintenance/

By the way... I started implementing support for KIF
syntax
  http://www.w3.org/2000/10/n3/KIFSyntax.py
but I have't gotten as far with it as I have with
the notation3 tools. In partcular, I haven't
done any KIF namespace stuff; maybe I'll give
that a try.

> Pat (Offline most of next week)
> 
> PS. If I understand tagging, it it needn't involve reification at
> all. You don't need to *describe* a sentence in order to attach a tag
> to it, you can just kind of point at it by ostention.

Can you elaborate on that a bit?

> And in any
> case, reification doesnt get you into the
> sentence/assertion-of-the-sentence distinction, which I think is
> where one wants to be here. (Maybe I don't really understand tagging,
> of course.)

me neither.

-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/
office: tel:+1-913-491-0501
pager: mailto:[email protected]
  (put return phone number in from/subject)


This archive was generated by hypermail 2.1.4 : 04/02/02 EST