Part of the Workshop Session Notes


Rules Workshop -- Session 3 -- Candidate Technologies 1

13 May 2005


Christian de Sainte Marie
Sandro Hawke



Speaker == Jos de Bruijn


Jos: WSML distinguishes between Conceptual Modeling (ontologies) and Logical Expressions

em, did you want that power strip?

<em> i'm swapping with Danny so I'm ok after all

<em> let me know when you need me to scribe

after the break is fine.

<em> ok

(ie Standards session)

<em> ok

<em> ... is there someone else?

RIght. DanC was moved to before lunch, in Massimo's absense

So we're ending early, with an optional TimBL-on-SemWeb session afterwords

<anthonyf> I would still like to have a voluntary demos opportunity ...

MikeUschold: what do you mean, separation conceptual model from logical expressions?

Jos: It's just the ontology is written in a different style. It's translated into a logical language. It's syntactic sugar.

TerryHalpin: On Marriage slide -- no ability to name parameters?

Jos: Just model it as a concept. We talked about naming parameters, but decided it would be too confusing with property names.

Terry: This is good way to handle this?

PascalH,Jos: WSML-DL is OWL-DL without Nominals (SHiQ)


HaroldBoley: RuleML


BenjaminG: (taking over on RuleML slides)
... Technical Scope of WG?
... Expressiveness: LP: Datalog-Horn + NAF + logical functions
... (logical functions do cause potential blow up with forward reasoning)
... [ THIS IS PURE PROLOG ] Dust settled 15 years ago.

<DanC_DCA> pure prolog includes NAF?

BenjaminG: Then enable use of RDF and OWL-DL
... And Draw on ...[all the other candidates]
... Then Add:
... - [long list, up to WG]

<DanC_DCA> "add some subset of the following features" <- eek! anybody know how to manage such a list? I don't think it can be delegated to a group. I think the charter has to pick. RDF Data Access charter didn't pick, and the WG is taking quite some time to put a cap on the feature set.

Q (next to Adrian) : Where does SWRL fit in....?

Benjamin: SWRL was an experiment. It kind of fits in to the lattice.

AnthonyF: Derivation <-> Reaction ? Validation?

<DanC_DCA> anthonyf: you mention Derivation and Reaction rules; what about validation rules?

Harold: Integrity Constraints. Starting to work on it. Had done them as queries which should fail.

Benjamin: Global constraint can be done easily in FOL-RuleML.

AnthonyF: You don't want everything to follow from a contradiction.

Harold: Paraconsistency.

SWSL Rules

<DanC_DCA> "loosely coupled but semantically rigorous" <- nice bumper-sticker

<tlrDC> Grosof assigns right to workshop organizers to publish his slides.

Bijan: This is a hugely ambitious schedule for a WG. What makes you think this is realistic?

Benj: This is what I was told was okay resource-wise, and I think it's possible.

MikeUschold: What syntax?

Benj: New sharded SWSL/RuleML-Presentation....


PhilArcher: Let me rephrase that: who would like to go to a seminar on the Semantic Web with TimBL? [lots and lots of hands]

<DanC_DCA> Experience with N3, slides

<DanC_DCA> TBL: somebody asked "how do you build real systems, when data comes from all over? how do you know what to trust?"

<DanC_DCA> ... one approach is to have it give you proofs and look at them, but another one is, rather than writing "if it's raining..." and hoping that it gets the answer from the national weather service rather than fred's email, you write "if the national weather service says it's raining..."

<DanC_DCA> that should be: { ?x fam:brother [ fam:son ?z] } => { ?x fam:nephew ?z }.

<DanC_DCA> (darn... I thought it would have been nice to go thru that example on the semantics/includes slide... I think it was a fairly interesting business rule)

Alas, he'd just gotten the 5 minute warning.

<DanC_DCA> ah... then he chose well

<DanC_DCA> this car example makes the relevant point

Dieter: Why not just make this the standard? Why a WG?

TimBL: some missing features. .... and of course Consensus.

DanC: Proofs?

TimBL: Yes, but it's broken

<DanC_DCA> 2 questions: Dieter said his serious question was: what's missing? he phrased it humorously as "why call a WG? aren't you done?"

TimBL: old proof stuff included verifier, that's lagged other changes.

Bijan: SW Requirements
... Conform to Abstract Principles (URI, open world, XML, distributability, layering)
... Conform to Existing Standards (RDF, OWL, SPARQL)

Four approaches to OWL Compatibility.....

scribe: Four approaches to OWL Compatibility.....
... 1. Subset (DLP, HORN-SHIQ)
... 2. Hybrid (Carin, DL Safe, AL-Log)
... 3. Superset (SWRL)
... 4. Alter (classic style rules, other non-mon extensions)
... 1-3 == subsumed by SWRL, so standardize SWRL

<DanC_DCA> hmm... yeah, AL-Log looks interesting.

scribe: AL-Log: looks a lot like prolog, with unary predicates handled by DL reasoner
... Carin: includes some binary predicates

<DanC_DCA> yeah, when you add forAll, encoding in triples is no longer practical

<DanC_DCA> (but I don't see why not allow subproperty of rdf:type)

Peter: Each of you has proposed or advocated a candidate, can you provide enough information for me to implement an interoperable .... IE full syntax and semantics.

Bijan: Yes. SWRL Submission

Ed: Do you mean *complete* ? Same answer?

<DanC_DCA> actually, PFPS gave 2 options: either enough to build an interoperable implementation, *or* a complete syntax+semantics spec

Jos: Syntax and Semantic on web, yes

N3: tutorial and test case. [ Peter: you failed ] [ Dan: No, test cases ARE enough ]

BenG: n3 last formal semantics
... n3 lacks formal semantics

<DanC_DCA> (test cases and docs)

BenG: SWRL is not well studied as a subset -- undecidable
... SWI Prolog, XSB, etc, ...
... RuleML presentation syntax on web; SWSL syntax same

<DanC_DCA> (I'll put the technologies deployed by docs+tests against the technologies by way of complete syntax and semantics any day.)

BenG: The Horn+Functions+NAF: Yes

Harold: with Well-Founded semantics

BenG: and including OWL is okay, too.

Harold: AL-Log reminded me of Sorted Datalog

Ian: "DatalogHorn+NAF = Tractable" ?

Beng: If you impose the rule that every number has a bounded number of distinct variables, then .... total comp. complexity is polynomial.
... the "VB" restriction

Yaser: For N3 -- is there a reason not to use named graphes instead of curly braces?

Tim: We don't have named strings, named integers, etc.
... you can just say x sameas { ... }

Harold: avoid named graphs, to avoid circularity

YaserBishr: We rely on the ordering of the list

TimBL: see Smalltalk's "Marriage of _ to _ on _" is the name of the function. Could do something like that.

HarryHalpin: Only Tim said Defaults and NAF wont work on the open web [ Bijan ack'd Open World ]; RuleML is maybe XML syntax for transferring rules around?

Harold: There are closed sub-webs, eg all W3C Recs, all states. For those you can use NAF.

Tim: When you're looking at a closed set, it's just a mathematical function

BenG: I use a broader defn of SemWeb --- "semantics" on the Web.
... The web is NOT open -- it is a collection of stuff, on planet earth.
... any given stuff on the web has an owner
... The set of conclusions that a set of premise entails..... You need to close off your sets.
... There's a presumption in all Mon and Non-Mon work that you close off the world.
... Syntax for explicit scoping is useful, of course.
... I think there's a misconception here.

TimBL: global symbol space of URIs; if two people use the same URI, they should be talking about the same thing.

Harold: "Predicate-Oriented Closed World Assumption"

Bijan; Is that equiv to circumscription?

Harold: No

<DanC_DCA> (perhaps if we put the panelists in the 4 corners of the room, the rest of us will be able to hear them talk to each other ;-)

DanielSelman: How much work have you guys done on Commercial rule bases .... eg FannieMae

Bijan: Not a lot
... I have looked at WS-Policy, at least
... for BenG - yes.

Michael?: N3 looks like F-logic without brackets and without semantics

Tim: It's very simple, except it's got the { }

MichaelKifer: Use the $ for reification

DanC: NAF?

MKifer: so it;'s a subset

BenG: F-logic has somethig like [ ] bnodes

Harold: Why do you think notIncludes isn't NAF

TimBL: It's not about failure.... it'd be easy to do with builtins

BenG: a sort of NAF Flat NAF

Dieter: Local Closed World?

BenG: "flat" -- you don't do it iteratively

Tim: It's explicit

<DanC_DCA> flat what? anybody got a citation?

CSMA: workshop on NAF

em: Thanks for context, SPARQL, Bijan
... You've been involved in ontollgy editing software, and and impls. How long your SWRL impl take?

bijan: it's not full -- but a couple days....

<scribe> Scribe: Sandro Hawke

Summary of Action Items

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.122 (CVS log)
$Date: 2005/05/13 19:35:55 $