02 Jun 2009

See also: IRC log


jar, +1.216.445.aaaa, hhalpin, dbooth, alanr, Ht




<jrees> me too.

<jrees> mibbit

<jrees> mibbit is good

I throw some URIs out:


The media-type on the ontology is wrong.



OK, go for it henry!

<alanr> so we can fight more clearly about it

<jrees> hh: Looking for big errors.

2 modules


In which we're trying to model in particular hayes vs. timbl debate

then ldow module

<alanr> what is the goal of modularization?

in which we're trying to model the linked data tutorial.

<jrees> hayes vs. timbl = accessible (on the web) vs. generic resource (not necessarily on the web)

<jrees> ?

Slide 7

Non-Information resource

and Associated Description

Formal webarch or IETF RFCs.

<jrees> hhalpin, your audio is dropping out occasionally

<jrees> modules = allow some communities to reject some modules.

slide 9 comment?

Is xsd:anyURI not a "URI?"

<jrees> alanr: xsd: anyURI is not a restriction of string

Does that reasoning make sense?

<dbooth> Alan: Comment on slide 9: xsd:anyURI isn't a string. Also not sure why you have xsd:anyURI. Is that so that you can make it a subject of a statment? Harry: yes.

OK, hasURIStrong -> hasURI?

At least one.


<Valentina> hi there

<Valentina> we are in a very noisy room

<aldogan> hi, pretty crowdy here, will try with the chat

<dbooth> Alan: Also URI to xsd:anyURI thing should also be inverse function, i.e., 1-1.

<jrees> dbooth, are you willing to scribe?

co-referential URIs

Two URIs that "identify" the same thing

That would be done on the level of resource.

with an "identify" property

But you could URIs that redirect to each other.

We could separate that out.

<Valentina> will try to call from skype and keep muted


asking owner

can owner via whois

<jrees> identifies - to falsify, you ask the owner.


this is a position :)

that's a position.

TimBL does.

We don't have "thing"

"thing" can be either a class or an individual

<jrees> alanr, that's a detail...

refers to catches Hayes's point

Hayes whole point really can't falsify "refersTo"

<jrees> how would you falsify refersTo ?

<jrees> provenance issue

john [ex:a irw:refersTo ex:orange] while mary [ex:a irw:refersTo ex:apples]

<jrees> john: A refersTo Oranges. mary: A refersTo Apples.

you can't really do provenance in straight RDF vanilla.

ex: uri irw:refersTo ex:PatHayesPerson
... uri irw:refersTo ex:PatHayesWebPage
... uri irw:identies ex:PatHayesWebPage WOULD BE WRONG

<jrees> hh: URI P refers to Pat, URI P refers to the web page.

<aldogan> ehm if you want to falsify, you need to include the context of provenance, can be made with a quad

<aldogan> however this is an issue for any rdf triple, not just refersTo I guess

it is a bit of level-mixing.

RDF does not really constrain your interpretation at all.

level of inference weak.

there are some objects/resources

<aldogan> agree

that are "real-world" things.

and, while formally, we can't restrain reference to refer to JUST this set of things

I want to communicate to someone

that my URIs refer to at least some of these things.

so that's the use-case.

That's also assuming Hayesian line.

<Valentina> I think that IRW should cover core concepts, specific modules should extend it for addressing applications such as trust, http transaction validation, etc.

<jrees> I don't even know what refersTo is meant to be from the 'logicist' point of view


That's fine.

But we're trying model this here.


<jrees> refersTo is not functional.

I think that was Pat's point.

<Valentina> yes, indeed

<alanr> informationrealization.owl has junk at the end that causes parse errors

<aldogan> because a resource can refer in too many ways

sorry, that's how tarkian models work.

<aldogan> oops, will control right now alan


I don't think that's right.

<jrees> what constraints are there on the interpretation of refersTo? is it functional?

<alanr> why is a awww:representation a non-information resource?

specifically meta-modelling interpretations

<jrees> x refersTo y if there is a model in which x is interpreted to be y?

allowing people to communcicate what interpretations their OK with.

they want people to know about.

<alanr> if irw:Resource is the same as rdfs:Resource, then why do you create it?

refersTo and identify are both controversial insofar as they are actually doing a bit of meta-modelling.


<jrees> is refersTo according to *one* model (chosen somehow), or *some* model?

model a set of infinite pairs

can be the real world

database model using semi-rings


<jrees> let's table this

<jrees> it's 9:30

<alanr> irw:WebServer–  A server hosts web representations. ? How does it host something that is "on the wire"?


operational semantics

<dbooth> I think RDF semantics is quite clear about what it means by "intepretation".

suspecious of this.

<alanr> irw:isResolution of talks about mapping to IP. How can that work? Hostname is essential for web transactions

<jrees> I don't think it's as complex as doing comprehensive metamodeling

<aldogan> alan, there were some characters that don't know how they fit in, maybe some server-issue, now updated a clean version

<Valentina> we model irw:Resource because we want to stay within OWL DL

<alanr> irw:isLocationOf english description doesn't match name

<jrees> Tim screams against a class of noninformation resources

that's we've moved that class to ldow module at this point

<Valentina> the class is annotated and that is explained there

<jrees> tbd: move noninforesource to LDOW so that it can be ignored by Tim.

<ht> HST objects to a class of resources, from which it follows that there is no _class_ of non-information resources

<Valentina> for me at least, such kind of feedback would be easy to be analyzed offline. In this way I don't know who and to what answer before, sorry

<Valentina> already did actually, I didn't put the new version online as here the connection sucks

<aldogan> if Pat wants to use refersTo with a strict model-theoretic interpretation, we need to provide appropriate extensions for answering questions like jonathan's ones

<jrees> info res. is broader than web pages, e.g. books

<aldogan> refersTo can be very general, in order to encode any reference, even non-tarskian

<aldogan> when I say "mary is very nice" I can refer to entities that can be well outside any strict model-theoretical interpretation

doesn't match other definitions.

Again, this ontology lets you state things.

<jrees> is beethoven's 7th an info resource by this definition?

If you believe that beethoven's information resource you should be state this vocabulary.

that fact using this vocabulary.

<jrees> hh: independent - answer to that question is not constrained by this ontology.

<aldogan> beethoven's 7th *score* is an information resource

<aldogan> distinguished from its performance

the particular performance of the score would definitely be an information realization.

<aldogan> right

<jrees> what is the 7th about?

if that particular performance as as an mp3 file over HTTP then that particular performance is Web repsentation

Notice that "isAbout"

optional not requirement.


<jrees> alanr: The score is a set of instructions... it's about playing the music...

<aldogan> an information resource is not about states

<ht> "Moby Dick" isn't 'about' any one thing. . .

<aldogan> ehm can be "isAbout" a state

foaf: depicts a subproperty isAbout

<jrees> alanr: If you can't answer questions like these, you'll be in trouble [due to inconsitent curation]

<aldogan> but an emotional state would be a real world stuff, not an information resource, at least in general intuition

<aldogan> we can answer questions like these jonathan

<aldogan> unfortunately i cannot speak from here

<aldogan> ehm alan, not jonathan sorry

<Valentina> isAbout is not funcitonal!

<jrees> alanr: You haven't given criteria

<Valentina> criteria?

logical positivist

logical atomist

<jrees> hh: These things are fundamentally subjective. You just can't say.

<aldogan> the score is the main way to execute the 7th

<aldogan> but does not sound, it is information

<jrees> hh: Positivism considered wrong by large amounts of 20th century philosophy

<aldogan> it is information that expresses instructions (content)

Read any Quine on Rudolf Carnap

<Valentina> if we enter such a philosophical discussion this will remain an ever ending story!

<aldogan> agree with harry

<jrees> hh: This is not a scientific closed domain. Falsifiability is not desirable here.

<aldogan> the ontology started from web science requirements

<aldogan> not philosophical ones

<jrees> alanr: How can you have meaningful communication if assertions aren't falsifiable?

<Valentina> can you please give me a clear example of falsifiability? sorry I missed the first part

<jrees> hh: It's not necessary.

<Valentina> but that's true for everything expressed in RDF

<aldogan> falsification can only be carried out within some specific assumption about the formal interpretation of a domain

<alanr> define red as luminosity over a certain range of the spectrum

"message" might be a kind of equivalent information realization

<alanr> blue as another

<alanr> if I say it is red and spectrum says otherwise it is false

<aldogan> so firstly, what kind of open/closed world assumption do exist on the web?

<Valentina> you are talking about consistency?

<aldogan> red and blue can be made different (or disjoint if classes) only in a controlled dimension

<alanr> if something is falsifiable, it is possible to make inconsistent statements.

<alanr> inconsistency is your friend

<aldogan> it applies well because you have the criteria to falsify in that particular context

<alanr> a theory should be good enough to be wrong

<Valentina> you cannot control consistency over the web

<aldogan> unfortunately, most of realistic contexts in real life are not as such

<aldogan> alan, you are acting like the first wittgenstein

<alanr> if the context is part of the truth conditions then so be it. this has nothing to do with whether something is falsifiable or not

<Valentina> we cannot prevent semantic web users to state what they want.

<aldogan> what you cannot "talk about" (=falsify for you) should remain silent

that's where at least to me the use-cases come from.

<alanr> anyone can say what they want

which this modelling of redirection

<aldogan> but the context of web reference is open

<alanr> that doesn't mean that everyone is right

<jrees> the difference is that alan is doing engineering, LW was doing something else

<alanr> yes, jar says I'm the anti wittgenstein

<aldogan> you cannot have *complete* interpretations for it

<aldogan> therefore we should live with limited falsifiability

<alanr> non sequitor

<aldogan> that depends, epistemically speaking, anyone could be right

<jrees> the measure of success is effective curation practice and ability to integrate, not truth about how people communicate in practice

<alanr> in one's ontology one should be able to tell what is what. Whether it is practical to determine it is a different matter.

<alanr> epistemology versus ontology

<aldogan> if you want strict truth criteria, you should provide an independent way to reduce epistemic freedom

<aldogan> and this can be dome in science, partly

<aldogan> while the web is social world, not only science

<alanr> I want something that we can push the science of web with.

<alanr> theories that can't be wrong don't help that

<jrees> no audio

<Valentina> I think this is not a requirements for IRW, that has a much more modest goal. Represent core concepts of web and sem web architecture in order to enable certain applications, such as "trust"

<alanr> gotta make hard choices ;-)

<jrees> all waiting for hh audio...

<aldogan> ok alan, now this is clearer

<alanr> for trust you better have some way to be wrong otherwise you trust everybody to no effect

<alanr> or don't trust to no effect

<alanr> after all, if you can't be wrong, what's it matter if you trust or not?

<aldogan> you want something that can enforce "strong" ontology in domains where this is a desirable feature

<aldogan> and of course this is a good use case

<aldogan> but it is not the only use case

<alanr> this had nothing to do with enforcing anything. it has to do with speaking clearly

<hhalpin___> Alanr: I am not interested in the verificationist criteria

<jrees> still no audio

<alanr> you - as an author of an ontology

<hhalpin___> Alanr: I am interested in what people are trying to communicate with using this vocabulary

<hhalpin___> alanr: I don't even like the word "ontology" prefer word "vocabulary"

<alanr> me too. What are they trying to communicate?

<ht> Harry does indeed see himself as late Witt. these days, in terms of how names work

<dbooth> I think one thing we have identified is that there is some confusion about whether harry's "refersTo" property is functional, given that there may be multiple interpretations (in the RDF sense).

<Valentina> refersTo is *not* funcitonal

<aldogan> i agree with you, but expressing subjective reference is "speaking clearly", but in different manners from typical scientific discourse

<dbooth> valentina, is refersTo functional in a particular RDF interpretation?

<alanr> what does refers to me?

<aldogan> refersTo is non-functional just to be able to accept different interpretations

<hhalpin___> hmmmm

<hhalpin___> zakim is not letting me in for somean? weird reason

<hhalpin___> stat code not valid?

<aldogan> commonsensical interpretations as references

<alanr> A

<alanr> from

<alanr> Web

<dbooth> aldogan, that is not needed in RDF semantics

<alanr> ick. cut and paste doesn't work

<Valentina> what does it means? refersTo is defined in such an ontology, you can use it as you want!

<Valentina> it is not an RDF primitive!

<dbooth> in RDF semantics, a property can be functional, but there can still be different interpretations.

<Valentina> actually IRW is modelled in OWL DL

<aldogan> late wittgenstein was about allowing full power to local contexts ("linguistic games")

<Valentina> and functional has a precise semantics, at least in my knowledge

<aldogan> subjective reference can be valid and falsifiable in a restricted group or community

<alanr> ?

<aldogan> in that case, you can create a specialized refersTo for the local linguistic game

<alanr> that was re: refersto

<dbooth> aldogan, if you are not using RDF semantics then I am going to be totally lost.

<aldogan> i am using "formal semantics"

basically, would like feedback about ontologgy

<aldogan> the one mostly intended by semanticists

<jrees> i would like to see the formal semantics for refersTo.

<Valentina> refersTo formal semantics

<Valentina> at the moment is an OWL object property

<aldogan> aldo gangemi

<alanr> can a uri refer to both Pat the person and Pat the page in the same world?

<dbooth> I think we have to start with RDF semantics as a basis, otherwise I (and probably others) will be totally lost.

<Valentina> with URI as domain

<alanr> is your answer: yes, each in different contexts?

<Valentina> and irw:Resource as range

<aldogan> i think yes, but it cannot "identify" it

<dbooth> alanr, by "world" do you mean rdf interpretation?

<alanr> i mean our world (good enought gloss)

<aldogan> my context here is the formal semantics of OWL-DL as adopted in IRW

<dbooth> alanr, your world is different than mine :)

<alanr> really?

<alanr> your brain is different, but our worlds are certainly the same.

<alanr> but perhaps you use the word "world" differently that I and many do.

<aldogan> let's distinguish the world we live in, not bearing a formal description ab origine

<Valentina> sorry Aldo and I cannot speak :(

<Valentina> we are in a noisy place

<aldogan> from the formal worlds we use to make engineering things :)

if the AWWSW demands debates about interpretation structures or logical verificationism

everytime someone wants to communicate

the group will get nowhere fast.

In particular, alanr, logical verificationism was more or less killed by Quine in the 1950s.

With a modest revival around Kripke and Putnam's "Meaning of 'Meaning'"

Re fetishizing logical semantics

As Pat or anyone can tell you

the logical inferences do NOT constrain worlds.

And often the worlds are not clearly composed of logical individuals

<jrees> david: Best we can do is to try to understand what Harry was presenting, difficult with HH off audio.

See Strawson's book "Individuals" for an investigation of this, or Evan's "Vareity of Reference"

<alanr> harry, I can't debate all the -isms

I'm just saying, getting into these debates is rather useless.

Almost no ontology can satisfy this rather silly requirements

<ht> I suggest a necessary question is "Why write ontologies" if you are in the Late Witt camp

<Valentina> I agree with harry

So, therefore, if making one fulfill these requirements of verificationism or buying into a rather simple-minded picture of how formal semantics works

ARE requirements for talking

then I see no reason to talk.

<Valentina> this is quire far from the goal, at least IRW's

However, if we can get good talk about what kind of things people want to talk about

<jrees> alan: Having the English description convey the intention of the authors well would be a good thing. We might be able to help with this.

<aldogan> let's be constructive people

<aldogan> 1) identifies, functional as in Tim's sense

regardless of whether these things can be "verified" or "have formal semantics"

<aldogan> 2) refersTo, non-functional

then I'm more for talking.

<jrees> dbooth: Is the ontology "U refersTo PatHayes. U refersTo PatHayesWebPage." going to be consistent?

<Valentina> we have modeled information resource, URI, indentifies (functional), refersTo, and some other

Will try to dial-in

I am more interested in how we can map this to generic ontologies

and bits of HTTP that we have not modelled, like "http Entities"

<aldogan> 3) refersToTarskian, non-functional, but with some other properties (Alan?)

<Valentina> we don't care about consistency on a web scale

<dbooth> I think alanr asked a good question: Can a URI refer to both a person and a web page in the same world?

<Valentina> the idea is that we can use this in order to do simple reasoning stuff such as

<aldogan> yes david

<Valentina> lod self describing sources

<aldogan> but not U refersToTarskian PatHayes. U refersToTarskian PatHayesWebPage

<jrees> consistency is a quantity to be maximized. Never infinite, but bigger is better.

<Valentina> semantic validation of http transaction

<aldogan> anyone willing to recap on email?

<aldogan> bye

<Valentina> please, let us have all your feedback and I'll send an updated version of IRW

<Valentina> bye

<alanr> email?

<jrees> adjourned, but people staying on the call & chat maybe.

<Valentina> email is fine

<jrees> continue in email... & now

<dbooth> my feedback: 1. I think more clarity is needed about whether refersTo is functional. 2. I think it is important to be based on RDF semantics as a common starting point, otherwise it will be very hard to achieve a common understanding.

<Valentina> I don't see the point to get rid of OWL DL

<Valentina> do you mean we should use the reified RDF semantics by Pat?

<aldogan> identifies is already functional, why do you need another one?

<Valentina> refersTo is a sort of non-stable property

<alanr> phone crapped

<alanr> will call back in a sec

<jrees> ht: weak & strong ambiguity

<Valentina> it is meant to capture the fact that you can have different usage of it

<jrees> ht: weak = usage may evolve over time

<alanr> ok. it needs to say that more clearly

<Valentina> I mean, imagine identifies represents the relation between a URI and the identified thing stated by the creator of the URI

<aldogan> weakness can be a strength

<jrees> ht: strong = synchornically same person may mean different things at same time

<aldogan> it catches a different requirement that identifies

<aldogan> than

<Valentina> while over the semantic web, the same URI is used to "refer to" things

sorry, can't skype on

<Valentina> hopefully, most of them will be same as of the identified thing

anyways, I would like actual substantive feedback

<Valentina> but don't know and cannot control

<aldogan> such reference was not intended to be model-theoretical

<Valentina> exactly

rather than bickering about falsifiability and misunderstandings of model-theory

we have use-cases

as detailed in the paper

<jrees> alanr: An instance of his refersTo relation would apply every time someone uses a URI to mean something?

<Valentina> yes

<Valentina> you

<Valentina> who are you?

<Valentina> :D

including dealing with HTTP-in-RDF.

And Linked Data

<aldogan> fully agree

Including letting one simulate 303 redirection

within RDF.

<Valentina> can you say that again?

<aldogan> also generic ontology of resources of tim should be integrated, it catches one of these very important use cases

which is useful for web-spiders, linked data, and that sort of thing.

<jrees> alanr: Then the ontology ought to have an entity that engages in the process of referring

So, I'm willing to try this again, but next time being more clear about ground-rules, lest we descend into insanity, which clearly is kinda what happened this time a bit.

We can an entity engaged in process of referring.

<alanr> the name should be "hasBeenUsedToReferTo"

<aldogan> :)

Although, I might add, some people don't think reference requires an entity that "refers"

<jrees> If someone adopts late W position, why write an ontology?

<alanr> doesn't matter what some people think

<alanr> we are working on *your* ontology

<Valentina> ok jrees, you mean we should have a special entity for referred things?

but more that it requires an entity that accomplishes some goal in lieu of a historical causal chain which accomplishes the goal due to some previous relationship.

<jrees> It's in order to play more satisfying language-games.

<aldogan> if the producer of the reference has to be engaged, we need to create a ternary relation

But yes, we can add an entity for a referring an agent though.

<aldogan> or to use a quad

I would not want that to be "required" though.

<aldogan> or to use some hack

<aldogan> but owl has only binary relations

<alanr> not owl2

<aldogan> we stayed simple there

<jrees> Satisfaction = not just warm fuzzies, but better health, more wealth, etc.

jrees - you can consider machine-languages language games for communication, and that can accomplish things.

<alanr> http://www.w3.org/2007/OWL/wiki/Syntax#Annotations

<Valentina> please, let me understand. I made an ex. the creator of a URI is the one who commits to the "identifies"

<jrees> ht: LW position is you can't do better than "it works"

jrees - there is nothing in late W that restricts that viewpoint to "natural languages"

<Valentina> you are suggesting we need to track the provenance of refersTo?

<jrees> I agree, hh

I would suggest we not use OWL2-specific features unless really required.

<aldogan> so do i

<jrees> We are arguing about the best way to make effective language-games (computer-assisted)

If we want, we already have the irw:Agent class

<jrees> Empiricism is a tool that I like to use.

<aldogan> refersTo is commented in that way however

<Valentina> did you look the "refersTo" rdfs:comment

<Valentina> I removed it

<Valentina> If I understand well

<Valentina> agent has been removed

<Valentina> ok

And we can say that an irw:Agent irw:referentialUse irw:URI irw:refersTo irw:Resource

This could present problems with graph merge though :(

<Valentina> sorry, not so explanatory

<Valentina> wait wait

Since multiple agents can be using the same URI to refer to different things.

<Valentina> ok, but in this case we would need a ternary

<Valentina> ok

I would suggest that we provide an ontology pattern

<Valentina> we are assuming people out of this

<Valentina> sorry agent

<Valentina> say that again

<Valentina> no

<Valentina> reference is not objective

but *not* make it required.

<jrees> Reference is not objective. Someone uses the URI to refer to something.

<Valentina> that's why it is not funcitonal

<aldogan> a ternary is viable however, i strongly favor the creation of a class irw:referentialIntention or something similar

How would you use that aldo?

just in brief in IRC?

<aldogan> by adding properties between it and an agent, and two resources

<Valentina> ok, tell me if this captures what you are saying: we need to express that a person/agent refers to something, through a URI

<jrees> When would one write an X refersTo Y triple, in practice?

<aldogan> this is the n-ary logical pattern

<Valentina> is that right?

<Valentina> in this case we need an n-ary relation

My feeling is that named graphs is the way to go with this sort of thing rather than model n-ary relations.

<aldogan> there is a tradeoff jonathan

but if people want to, we should provide them some ontology design pattern to do so.

<jrees> Don't need to model n-ary... just need to be clear where the quantifier is.

<aldogan> one triple is easy, but has limitations because you need hacks to include provenance

<Valentina> because you are talking about a relation between three arguments

<Valentina> agent, uri, thing

<jrees> three-place relation agent uses URI to refer to thing

<aldogan> three triples are difficult, but include provenance information

<Valentina> refersTo(Agent, URI, Resource)

<aldogan> yes, sorry aganet-

<aldogan> agent-uri-thing

<jrees> deal with this, you can reify; you can do some ternary-to-binary reduction; or you can quantify

<alanr> process of a certain type, with participants: agent, name, thing.

<aldogan> exactly

<jrees> e.g. *there is* an agent that uses URI to refer to thing.

<aldogan> yes, we have a pattern called situation: http://www.ontologydesignpatterns.org/cp/owl/situation.owl

<aldogan> for that alan

<Valentina> that's fine, we just didn't want to model agent in IRW, but we can put it in again (it was there actually)

<aldogan> the original IRE had that style of representation

<alanr> ok, then use it

<Valentina> ok, I'll fix it and propose you an updated version

<Valentina> I am fine with Agent in

<aldogan> better an alternative, to be submitted to vote

I do not think we should *require* that design pattern, but allow it and provide examples.

<Valentina> require?

<jrees> what is a requirement?

As in, disallow ex:a irw:refersTo ex:x, irw:refersTo ex:y.

<jrees> the definition is up to you. I would just like to know what you mean

I think I should be able to say that inside a named graph, or even just put that in RDFa inside my web-page.

<dbooth> jrees, are you again asking if irw:refersTo is functional?

<Valentina> I think that we have clarified now refersTo

<jrees> refersTo1 = *someone* used the URI to refer to the thing

again, it is *important* for me to make sure irw:refersTo is not functional. If you wish, you can use irw:identifies, which is functional.

<aldogan> consider that there are interaction aspects there

<Valentina> we meant a URI refersTo a Resource (and we kept implicit who stated such a triple)

<aldogan> imagine using a situation insetad of a property to add an RDFa tag ...

<jrees> refersTo2 = *some particular agent, inferred from context* used the URI to refer to the thing

<Valentina> with identifies we meant a URI identifies only one thing (give by the creator I would say)

<Valentina> what we did is to get rid of Agent in IRW

<aldogan> rel:AldoIntendingToStateReference

<Valentina> and kept it implicit

<aldogan> ulp!

<alanr> refersto1 : exist t st. exists person, person used uri to identify thing at t

<Valentina> we can make it step in again

<jrees> in refersTo2 = that agent might be the author of the RDF graph, maybe.

I mean, if we're going to do this for refersTo, we might want to this for identifies

<jrees> refersTo2 = functional, refersTo1 is not.

<dbooth> oh. so irw:refersTo is not functional, but irw:identifies *is* functional?

as irw:identifies in TimBL's story identifies because the "owner" (an agent) wants it to identify.

Yes dbooth - irw:identifies is functional.

<Valentina> ok

<aldogan> alan, do you mean that refersTo has to defined in terms of identifies?

<Valentina> but you're mixing terminology

<aldogan> dacid, we have said that many times :) and is in the owl

<jrees> if I say "U refers to X" the question according to whom?

<alanr> doesn't have to. I'm trying to understand what you are saying and it seems like that is what you are saying

<jrees> possible answers: someone, everyone, Harry, or ...

<Valentina> somebody on the sem web

<aldogan> no, we weren't say that

<jrees> weren't saying what?

<aldogan> an agent may want to state a non-exclusive reference

<Valentina> "U refers to X" according to somebody on the web

<jrees> that's refersTo1

<Valentina> ok yes

<aldogan> for example that http://www.w3.org refersTo W3C, a web page, and a legal person

<Valentina> this is "irw:refersTo"

I believe you can say "U refers to X" does not have to clearly have an interpreting agent.

<Valentina> "U identifies X"

<Valentina> according to the creator

Often that is implicit, and there are theories of refernence where that is not required.

however, I do think allowing people to say "according to whom" is useful.

<aldogan> the case you catch with your axiom is when an agent refers to st at some time, exclusively

<jrees> to falsify refersTo1, you check every agent, and if no agent uses the URI in that way, then the statement is false.

Every agent at all points in time :)

<Valentina> can I suggest we have a solution for at least such two properties

<Valentina> ?

<Valentina> we put in Agent

<Valentina> in

<alanr> it ok to do that, harry if it helps people understand what you mean

<jrees> refersTo3 = all agents use U to refer to X (all the time?)

<Valentina> and I fix the English description including examples!

<Valentina> all?

<alanr> yes. always include examples of usage

<aldogan> ok we need to provide an alternate version that takes back in some original IRE classes

<jrees> alanr, to falsify refersTo1, you would have to monitor all agents for all time... ergo not falsifiable.

<Valentina> refersTo3, I don't agree it's needed

<Valentina> I mean

alanr - I'm just saying, read Quine's rather famous "radical translation" argument here, the problems on verifying even simple language use are rather deep, and get worse rather than easier the further one is from scientific discourse.

<jrees> refersTo3, never true, because I can always construct a nonparticipating agent

<Valentina> what do you mean with "all"?

<aldogan> jonathan to get that, we need a temporary close world assumption

<aldogan> closed

<aldogan> actually that is a practical solution while living with the simple triple solution

<dbooth> harry, can you give us a web pointer to Quine's rather famous "radical translation" argument?

<alanr> http://en.wikipedia.org/wiki/Radical_translation

<jrees> This is a language-game, and what we say has something to do with the world? So tell me how to play the game - when would you like me to use refersTo, and when not?

<aldogan> gavagai! :)

<aldogan> is quine's example for radical translation argument

<alanr> just because you don't know what gavagai means, doesn't mean it doesn't mean something very specific

alanr - yep, but you can *never* know with certainty, that's the point.

<aldogan> you may want to use it when you are stating some reference without having the authority to declare an exclusive identification

<aldogan> or

<alanr> so?

<aldogan> when you want to allow alternative references

<aldogan> or

<jrees> certainty is not the right crierion. we want usefulness

<aldogan> when you want to import references from other datasets

<Valentina> the point is that refersTo capture what actually happens

<alanr> moreover the thought experience is not active. It's not clear that unknown distinctions couldn't be figured out by some "game".

<Valentina> the fact that u use URIs and put them in triples

I do have sympathy for alanr's position, and I think something like "verifiability" does hold for certain scientific discourses, like chemistry. But in general, and since the Semantic Web captures "in the general", the point is we can't depend on verifiability.

<dbooth> I think "having the authority to declare an exclusive identification" misses the point. The problem is not whether someone has the *authority*, it is that there will still be different interpretations possible.

<jrees> so this goes to the indexicality (?) of RDF utterances... can RDF contain pronouns? according to rdf semantics, yes, but what would be the practical effect?

<Valentina> this may lead to discover u are using that URI for referring something that is not the thing it was created for

<Valentina> or

<alanr> whether something *is* verified is different from whether something is *possible* to be verified.

<aldogan> harry, that was my point before

<Valentina> maybe it is something sameAs the thing it was created for

<jrees> declarative vs. imperative maybe? the semweb is constructed, so maybe its assertions are imperative. science is discovered, so maybe its assertions are declarative (falsifiable). is that sensible?

<aldogan> possibility of verification is granted by several strategies

<alanr> no

<alanr> to jonathan

<aldogan> "actual" (?) verification is too heavy a requirement to me on the web

<aldogan> btw it is caught through functional "identifies"

<alanr> for example, gavagai. Suppose you can communicate that you want the speaker to test your knowledge.

<aldogan> or by some subproperty of refersTo

<jrees> I'm trying to understand the IRW philosophy, that leads to definitions of the sorts that it makes.

alanr - we are not denying the possibility that *some* statements on the SemWeb (like those related to health science!) are unverifiable, we're just saying not all are.

<alanr> He says a bunch of things about gavagai and you look at him blankly or pick up the tail instead of the leg.

<alanr> then you don't understand

<jrees> so that it can be given a fair shake and compared to, say, OBOF.

<Valentina> sorry I will go now, have to finalize slides for tomorrow presentation of IRW :)

<alanr> we're at the stage of establishing ground rules, I think

<alanr> I don't think OBOF are only ones possible

<aldogan> IRW philosophy is partly based on IRE, which was built having semiotics in mind

<Valentina> thanks for the discussion, will continue by email

<jrees> we are? does hh agree?

the problem is that are unresolvable differences re ground rules.

<jrees> i guess he doesn't.

<alanr> no, there are only busy philosophers

and everytime I try to interact with this group I get into far more disagreement re ground rules

<jrees> why are they unresolvable?

than practical engineering.

because if you believe everything is verifiable, that's fine by me.

I think that's wrong.

My goal is not to convince alanr that logical verificationism is wrong

<alanr> harry, we are very practical engineers. Look what we build.

<jrees> I want to understand your position, hh, and don't

<alanr> everything I do I do because I've been screwed by doing something different

<jrees> I think alan is a bit of a bully

<jrees> sometimes

<alanr> (sheepish)

or to try to convince dbooth that Tarski-style interpretations have almost nothing to do with the world besides constraining inferences.

<dbooth> harry, will you take a look at http://dbooth.org/2009/denotation/ and tell me your thoughts?

ok, will do.

<aldogan> gavagai: i pick up something wrong, ok, so what? i have been told that refersTo is one viewpoint, not the truth

<dbooth> thanks

but i would like an ontology for resources.

i would like that, before, say the end of the year.

<alanr> aldo: an alternate definition is fine.

the poor HTTP-in-RDF people need one.

<jrees> tarski's big, what's the reference? do you mean model theory?

<aldogan> we all use heuristics from our previous knowledge, even when reading OWL

Linked Data could use one.

<alanr> Harry is arguing about the impossibility of giving a definition

Web-spiders could use one.

<aldogan> without them, not much can be understood

<alanr> by citing gavagai

<jrees> do we compete or cooperate?

Thus, we stop trying to argue over ground rules and try to model what we can.

<alanr> and I am questioning how carefully he is thinking about it (he says, indelicately)

Even if we don't agree.

about metaphysics.

<aldogan> first, consider that owl is not appropriate to represent expressive axioms

<jrees> then how about goals.

<alanr> harry, we're trying that too.

I might add no-one belives in logical verificationism anymore, alanr, by the way.

<alanr> oh, I'm stung!

And the criticisms you direct at irw could be directed at any ontology.

<alanr> And I do

about almost anything.

<alanr> about the ontologies I build as well

And thus, I refuse to argue about nonsensical objections based on mistaken philosophical positions.

<alanr> and where I miss, Jonathan kicks in

<alanr> but you don't present engineering cases either

Yes, but we do.

See end of the slides.

That's a great caes.

<jrees> I see alan's approach as a tool, a subroutine or technique toward a goal

We can say "look, we can model 303 redirection, and determine whether or not something is a non-information resource"

purely delcaratively without HTTP mechanics.

And the people from Freebase and others at Vocamps liked that.

<alanr> foo type noninformationresource

<alanr> basic rdfs

<aldogan> sorry battery exhausted, lost last 3 minutes

As their main objection to Linked Data/TAG talk was that it was crazy to require HTTP mechanics like 303 all the time.

<dbooth> I do think the 303 redirection use case is important to be able to model, but I think it is largely independent of this whole question of identity and reference and what is an information resource.

So, I'm happy to talk about the ontology more.

Would like to see reviews.

<jrees> yes.

Saying "hey, this property's definition is confusing" or "this class is wrong" or "please add this constraint"

<jrees> goal: "help the HTTP-in-RDF people"?

<alanr> k, take care.Let us know how the presentation goes

However, arguing verificationism is not what I want to do with my spare time.

<aldogan> eheh

<jrees> tell me success criteria then, or something else I can sink teeth into

I hope that is understandable. If others want to argue ground-rules, that's fine.

<alanr> harry, remember I did say that. 1) Audit for circular definitions 2) include examples of usage of each term 3) on terms I or Jar said we didn't understand, think hard about restating.

<dbooth> harry, my main comment so far is that i don't understand why "refersTo" is not functional, given that in RDF semantics a functional property can still have multiple interpretations anyway.

Success criteria: have an ontology of resources that others can use that fulfill some use-cases

Put it up as a SWIG note or something.

So people can find it.

That's simple.

<alanr> david: because he means different people using the same uri at different times to mean different things

<jrees> use cases being...?

<alanr> all the things that were meant are the object of refers to.

<alanr> they are collected on that property

dbooth - then you can's state "a irw:refersTo b, irw:refers c. c disjoint b"


<aldogan> david, separate thread for that

<aldogan> if you put something functional, you allow only a single individual interpretation when you use that property monotonically

we need to be able to state that.

<aldogan> and this is not what we wanted to catch, because it is already with identifies

Alan, I disagree with 1), am OK for 2), and maybe on 3).

Example-driven is way to go.

<alanr> is "a irw:refersTo b, irw:refers c. c disjoint b" something you want to say?


that's fine.

<alanr> I have no idea what it means

<alanr> disjoint is a relation between classes

<alanr> refersto relates individuals

i think www.google.com refers to a web-page and a company.

<aldogan> ok alan, thanks for feedback

I think companies and web-pages are different.

<alanr> but it makes no sense that a web page and a company are disjoint

And I can use ONE URI to refer to both.

I did not say owl:disjointWith

<alanr> yes, that's your definition. I understand it now.

<aldogan> maybe harry wanted to say "a irw:refersTo b, irw:refers c. c differentFrom b"

<aldogan> owl:differentFrom

<dbooth> ok, so maybe they want to say "a irw:refersTo b, irw:refers c. c != b . "

That we can have multiple interpretations, and b and c can satisfy them, and life is ok.

<alanr> not sure why you want to say that, but at least it is well formed

<aldogan> it is well formed also the following:

<alanr> consider: a irw:refersTo b; irw:refersTo b. Conclude b!=b

<aldogan> a irw:refersTo b, irw:refers c

<alanr> no unique name assumption

<aldogan> alan is right, we do not want to state that b and c "must" be different

<aldogan> they could be, or not

<jrees> the question again is, how must we interpret 'disjoint' (according to the rules set out)? can one exhibit a single model that's consistent?

<aldogan> it is a relaxed way to represent reference

no, but they could be different.

that's fine.

<aldogan> a irw:refersTo b, irw:refers c, conclude i do not know if b=b or b!=b

<aldogan> like in many cases in everyday life

you can construct a single model and have MANY things satisfy that model.

now, you could try to reify interpretations

but I'd prefer not to at this point.

<alanr> got to head out now. Take care..

<jrees> any time you take meta-properties (like refersTo) and meta-classes and drop them down into an ontology you will have to work out your model theory... otherwise you're asking for trouble (nonsense, inconsistency)

<jrees> right, i have to go now too. we should adjourn to email. thanks all

<dbooth> so it sounds like "x irw:refersTo r" means that someone *said* that x denotes r.

<alanr> david, approximately. Actually someone used x to denote r, where denotation is in the eye of the beholder

<dbooth> which tells us nothing about whether x really *does* denote r.

<alanr> denotation isn't a property of a symbol, it's something that comes about when a person uses a symbol.

<dbooth> y

<alanr> our goal in communication is to try to make is so more than one person can use the symbol to denote the same thing.

<dbooth> well, if you're modeling the symbol explicitly then it *is* a property of the symbol though.

<alanr> (one goal)

<alanr> nope

<alanr> it's a property of both

<alanr> unless you define a different property that removes one of the things related by quantifying over it.

<dbooth> I'm not sure that's our goal, unless we recognize that "same thing" may have varying interpretations. We do want to *constrain* the interpretations though, to be good enough.

<alanr> the words "same thing" can mean different things. but when I used the words I mean that they *are* the same thing.

<alanr> :)

<alanr> ok bye.

<dbooth> :)

quick note: the model theory of RDF does not say whether or not x really does denote r.

so, thus, reworking model theory is not necessary per se.

since RDF and its model theory say nothing about what satisfies the model.

interpretation constrains a bit, but not very much, especially with RDF(S). OWL a bit more.

take care, gotta work on slides.

<jrees> rrsagent. draft minutes

Summary of Action Items

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.135 (CVS log)
$Date: 2009/06/02 15:17:38 $

Scribe.perl diagnostic output

[Delete this section before finalizing the minutes.]
This is scribe.perl Revision: 1.135  of Date: 2009/03/02 03:52:20  
Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/

Guessing input format: RRSAgent_Text_Format (score 1.00)

Succeeded: s/aion/ation/
Succeeded: s/against/against a class of/
Succeeded: s/if/is/
Succeeded: s/me/mean?/
Succeeded: s/right/write/
Succeeded: s/Does/Is/
Succeeded: s/ird/irw/
No ScribeNick specified.  Guessing ScribeNick: hhalpin
Inferring Scribes: hhalpin

WARNING: No "Topic:" lines found.

Default Present: jar, +1.216.445.aaaa, hhalpin, dbooth, alanr, Ht
Present: jar +1.216.445.aaaa hhalpin dbooth alanr Ht

WARNING: No meeting title found!
You should specify the meeting title like this:
<dbooth> Meeting: Weekly Baking Club Meeting

WARNING: No meeting chair found!
You should specify the meeting chair like this:
<dbooth> Chair: dbooth

Got date from IRC log name: 02 Jun 2009
Guessing minutes URL: http://www.w3.org/2009/06/02-awwsw-minutes.html
People with action items: 

WARNING: Input appears to use implicit continuation lines.
You may need the "-implicitContinuations" option.

WARNING: No "Topic: ..." lines found!  
Resulting HTML may have an empty (invalid) <ol>...</ol>.

Explanation: "Topic: ..." lines are used to indicate the start of 
new discussion topics or agenda items, such as:
<dbooth> Topic: Review of Amy's report

[End of scribe.perl diagnostic output]