- From: Harry Halpin <hhalpin@ibiblio.org>
- Date: Tue, 04 Apr 2006 14:19:23 -1000
- To: adasal <adam.saltiel@gmail.com>
- Cc: Paul S Prueitt <psp@virtualtaos.net>, Danny Ayers <danny.ayers@gmail.com>, semantic-web@w3.org, Azamat <abdoul@cytanet.com.cy>
I was objecting to the discussion of UFOs because to me the idea seems
patently wrong, and the fact that the people arguing for it instead of
making their UFO in OWL-DL/RDF proceed to criticize the entire Semantic
Web enterprise, which to me seemed invalid on the basis that they have
nothing better to offer.
I do apologize if anyone is insulted, but since I know that most
Semantic Web people are not from a background in
(philosophy/psychology/logic) but are instead practical developer types
(hopefully, and that's a good thing!), then I just don't want people to
think UFOs are a good idea without knowing there is almost 40 years of
well-respected people saying they are a genuinely bad idea. And that
most foundational ontology\KR based companies minus Cyc more or less
crashed and burned in AI Winter, a sort of analogue of the dot.com
crash. Feel free to try again, although currently venture capitalists
are more interested in tagging than ontologies. I also feel like
occasionally UFO advocates pretend to represent the mainstream of
academic AI and philosophy, which is not the case.
Here's a quick list of references by well-respected people that support
the viewpoint that metaphysics or "common-sense" is not suitable for
formalization and the UFOs are a bad idea:
In philosophy, psychology and neuroscience, almost all work proves that
representation or formal thinking is a very useful skill that is domain
dependent and deployed in particular contexts using heuristics (much
like Newell & Simon thought..). The best easy-to-read source is Andy
Clark, particularly "Being There: Bringing Mind, Body and World Together
Again":
-> http://psyche.cs.monash.edu.au/v4/psyche-4-14-chemero.html
As regards metaphysics, Brian Cantwell Smith (former LISP compiler guy
that brought us the ideas of reflection and reification in his Ph.D.
thesis) has written a pretty convincing argument that there are no
metaphysical facts of the matter. See his "On the Origin of Objects"
-> http://mitpress.mit.edu/catalog/item/default.asp?ttype=2&tid=8167
Another good bit of writing Smith has is "The Owl and the Electric
Encyclopedia" in which he correctly notes that Cyc will never finish
formalizing "common-sense" and provides compelling reasons, such as that
even "simple metaphysical facts" are embodied and perspective-laden:
-> Smith, B. C. The Owl and The Electric Encyclopedia. Artificial
Intelligence 47, 1-3, Jan. 1991, pages 251-288 (1991).
Note that he knows from experience - he was on the team of KRL
(Knowledge Representation Language), a famous AI KR system by Bobrow and
Winograd. Due to its lack of a proper formal semantics, the system could
express anything but could not correctly pull off even simple inference:
-> Experience with *KRL*-0: One Cycle of a Knowledge Representation
Language. IJCAI 1977: 213-222
This led another generation researchers to investigate formal semantics
and the trade-off between tractable reasoning and expressive ability.
This gave us the field of description logic, which OWL-DL is made in.
This long and hard 15 year lesson is why the Semantic Web uses formal
semantics, description logic, and focuses on inference (and thus things
like making n-ary predicates hard to represent and why its better to
give someone "legos" to build an ontology out of rather than trying to
build a foundational one). The classic paper is Brachman's
-> H. Levesque, R. Brachman. Expressiveness and Tractability in
Knowledge Representation and Reasoning. Computational Intelligence, 3:
7893, 1987.
If after reading these things you still think it would be easy to
formalize space and time, please take a look at Pat Hayes's "Catalog of
Temporal Theories", which shows you there are many and incompatible ways
to represent time:
-> www.ihmc.us/users/phayes/docs/TimeCat96.pdf
If you believe that spatial relations are easy, please notice the great
case made against even using 3-d Cartesian co-ordinates from the
viewpoint of linguistics and psychology, since it appears that's not how
the brain or human language work at all as given by:
-> Spatial Prepositions: A Case Study from French by Claude Vandeloise
If you still think UFOs are a good idea after noting that mainstream
psychology, philosophy, and KR has more or less given up on them, you
can download and use the UFO Dolce-Lite, which is actually made using
OWL and is available publically for download:
-> http://www.loa-cnr.it/DOLCE.html
Also, looking at Wikipedia always helps:
http://en.wikipedia.org/wiki/Upper_ontology_(computer_science)
I might add very bright people like Guha (Cyc employee that helped
RDF-S) are no longer working on foundational ontologies, but working on
the Semantic Web, syndication, and now Google. Note that after the
failure of KRL, Winograd moved into the field of HCI and some of
students gave us...Google.
Cyc, the longest running effort, is still not working all that well :)
and philosophers can't decide if objects are static or processes (See
articles by Varzi as well as arguments over SUMO:
http://www.columbia.edu/~av72/recent.html). Lastly, note that Brian
Smith, a compiler expert turned metaphysics professor, when criticizing
the entire foundational ontology approach, correctly said the Web was a
much better idea:
"Forget intelligence completely, in other words; take the project as one
of constructing the world's largest hypertext system, with CYC
functioning as a radically improved (and active) counterpart for the
Dewey decimal system. Such a system might facilitate what numerous
projects are struggling to implement: reliable, content-based searching
and indexing schemes for massive textual databases."
-- Brian Smith, "The Owl and the Electric
Encyclopedia"
I just happen to agree and would replace "Cyc" with "decentralized
ontologies" or "statistical NLP" (what Google uses). The conflation of
UFOs with the SemWeb will doom the SemWeb, which could be genuinely
useful way to exchange and link data, to being a sort of ivory-tower
pie-in-the-sky much like most UFO projects such as Cyc that have given
AI a relatively bad name over the years. I also might add Sowa's
conceptual graphs are well-respected, but not exactly making him as much
money as Google :) KR people have a long history of hyping up UFOs and
then the UFOs proceeding to be impossible to formalize and not useful at
all - see "AI winter," where the US military withdrew tons of money from
AI research and KR-based companies went bankrupt. I just don't want the
SemWeb repeating the (somewhat large) historical mistakes of AI that UFO
advocates seem to want to jump right back into.
Note I have written this all down before as well if you want more
references and argumentation:
http://www.ibiblio.org/hhalpin/homepage/publications/html/airedux/
However, go ahead - download DOLCE or formalize your UFO - and then
e-mail me when you think Tim Bray would give you the RDF.net prize, or
it would achieve widespead and non-domain specific use.
cheers,
harry
adasal wrote:
> Seems etiquette is to reply to all.
> Anyway Azamat, I apologize for misspelling your name. Probably not the
> only mistake in my post -:)
> It seems like a huge amount of money, but one should understand that
> these are terms that governments talk to each other in. As you know
> the ostensible wisdom is free trade now days, this is back door state
> pump priming.
> What does seem to me to be relevant is that different approaches be
> represented in this effort. Actually I find it hard to believe
> (perhaps you know?) that in such a huge spend variations of approach
> such as you might advocate and, perhaps one John Sowa might have
> conceived, are not being explored.
>
> Well I actually owe Paul Prueitt an apology (he will know) as well,
> which is necessary before I make my next comments.
> Harry,
> First of all is there a semantic-web-philosophy@w3.org
> <mailto:semantic-web-philosophy@w3.org> list? I guess this
> conversation would belong there if there is.
> That said arguments are not won by force and bluster.
> I find this type of discussion useful to my work, but the very idea
> that it couldn't be smacks of coercion, the very thing you are riling
> against when you say this is ideology. I don't believe that is so for
> a minute. I frankly don't understand a to of the issues being
> discussed here, this is one of the reasons I find it useful, but with
> regard philosophy, specifically metaphysics, there is a powerful and
> cogent argument that there are such things as metaphysical facts. I
> don't believe that there is any attempt to indoctrinate with the one
> true ontology approach. But I do think there is an attempt to
> delineate the landscape in a clear way and that clarity will entail
> building on certain facts, yes, metaphysical if you will.
> As I have said, I don't understand the field well enough to really
> contribute. On the whole my attitude is if there is something that
> operates consistently on a higher level than RDF/OWL and I can get an
> intellectual purchase on it then, hopefully, so much the better for
> me. But anyway, in the mean time I am learning about the issues that
> surround RDF usage.
> I have to say that I strongly object to any attempt to curtail my
> participating in such discussions.
> When it comes to metaphysical fact I know that the sine qua non of a
> human is the ability to reason and that reason entails language. My
> thoughts and questions are fairly unbiased. Force majeure is not
> rational argument. But my foregoing comment is equally as critical of
> for instance this passage in the paper:-
> Analogical Reasoning <http://www.jfsowa.com/pubs/analog.htm> by John
> F. Sowa and Arun K. Majumdar
>
> According to Aristotle, the essence of human includes both
> rational and animal. Other attributes, such as laughing or being a
> featherless biped, might be unique to humans, but they are
> /accidental/ attributes that could be different without changing
> the essence. Ibn Taymiyya, however, maintained that the
> distinction between essence and accident was arbitrary. Human
> might just as well be defined as laughing animal, with rational as
> an accidental attribute.
>
>
> I, therefore, believe it is a fact that humans are rational in the
> sense that we can ascribe reasons to all thoughts, feelings and
> desires. It cannot, therefore be an accidental or arbitrary
> distinction. I am alerted to the possibility that analogic reasoning
> is, as its name suggests, in fact a form of rational process which
> cannot really be set against other forms of reasoning. It is a
> question as to what distinguishes it from other forms.
> The following is an illustration of how rational discourse works with
> respect to metaphysical fact by another obscure philosopher, Dr G.
> Klempner, in a discussion of Dummett and Wittgenstein:-
>
> From Dr Geoffrey Klempner Pathways Letters to My Students Michael
> Dummett on 'realism'
> <http://www.philosophypathways.com/letters/marg8.html>
>
> My vision of the so-called 'problems of metaphysics' consists in
> the recognition of the simultaneous necessity and impossibility of
> expressing, say, anti-realism or immaterialism. There are no
> theories that can be stated, or refuted, but only the dialectic,
> the continuous struggle to find an adequate expression. Or you can
> think of it as a battle against metaphysical illusions, where the
> illusion keeps finding different ways to 'manifest itself'. There
> is no such thing as 'realism', there is no such thing as 'anti-
> realism'; there are only the various things that, as a putative
> 'realist' or 'anti-realist' one is more or less irresistably
> tempted to say.
>
>
> This adds strength to the notion "no one true ontology", while not
> denying that there may be a very favourable way of exploring the logic
> of the field.
> What I am interested in here, in this forum, is some instrumental form
> of reasoning that would suit my pragmatic purposes. For this reason I
> am also interested in the economics and politics of it. I want to know
> if I can utilise what ever I know in an as yet under-represented field
> to my advantage. So far, this is not difficult to follow or
> understand. When I am told that there is work that is being got on
> with, obviously I know that is the case, but I should point out that,
> not with standing the huge spend in the area, there is another huge
> spend being made by the UK in Health IT (the largest such project in
> the world) and yet there is very, very little evidence that the
> benefits of one are being used to the advantage of the other. This is
> both alarming and astonishing.
> I face two ways on this. On the one hand I would like to see what
> could be made relevant to that field (I mean beyond HL7 and Dublin
> Core). On the other I have to seriously question if either the methods
> are correct, the tools appropriate or what it might be to account for
> this disjuncture.
> Best,
> Adam
>
>
> On 03/04/06, *Harry Halpin* <hhalpin@ibiblio.org
> <mailto:hhalpin@ibiblio.org>> wrote:
>
> You're right - I am missing the point - it's not philosophy. When
> arguing against people that believe in the "One Big Ontology" approach
> or the "Perfect Design" approach, it's an argument against ideology.
> It's like arguing about the existence of God, and about as
> productive. :)
>
> And I'm not arguing against pragmatics per se, I'm arguing against
> standardizing notions of pragmatics. Instead, by decentralizing the
> creation of ontologies and allowing people to expose data as they see
> fit, they will build off their concrete real-life situations and
> experiences. Over time, useful abstract ontologies may or may not
> emerge. And yes, it is messy making this stuff fit in
> RDF/OWL-DL/whatever, but the same would hold with any formalism,
> period.
> And what I am arguing is that if people are supposed to use in a
> decentralized manner to build out of one standard centralized ontology
> (i.e. Entity/Endurant/whatever) and one that tried to delimit
> pragmatics
> (illocutionary/some bizarre version of control theory/etc) then well,
> it's going to *a lot worse.* Give people minimal constraints, not
> maximal ones.
>
> So get around to expressing your pragmatics in OWL-DL or RDF or KIF or
> whatever. And then show a real-life use case. Then who knows, nothing
> prevents you from standardizing it yourself. Then if enough people
> use
> it, the ISO or W3C could give it their stamp of approval. But to
> critique the Semantic Web for not reading (Fill in your favorite
> ontology/philosopher here, like Bunge) and standardizing him is kinda
> silly, since that would obviously be a case of premature
> optimization.
>
> Until then, I'm going to do what I'm sure almost everyone else on this
> list-serv is doing, which is ignore this whole thread so I can get
> some
> work done :)
>
> Paul S Prueitt wrote:
> > You miss the point, as most do in the W3C column.
> >
> > For a discussion of the issue of representation of reality with
> a formal
> > system, please review
> >
> > http://www.bcngroup.org/area3/pprueitt/kmbook/Chapter2.htm
> <http://www.bcngroup.org/area3/pprueitt/kmbook/Chapter2.htm>
> >
> > And citations referenced ...
> >
> > It is not correct to think of this as philosophy. There are
> real practical
> > problems with the notion that formalism (created by knowledge
> engineering
> > individuals often without deep insight into domain specific
> context) would
> > be found acceptable outside of these (knowledge engineering
> context). BioPax
> > is perhaps the best example of good cell and gene signal
> expression ontology
> > - and this ontology is designed to take a step towards data
> sharing - not
> > designed to explain signal expression. Again, practical issues
> arise when
> > OWL is used in complex situations. One can work around this, as
> BioPAX
> > does; nicely, but one cannot remove certain issues (related to
> degeneracy of
> > entailment in specific instances).
> >
> > But it is not merely that the wrong community might be designing
> ontology
> > for the rest of us, it is that (any) formalism is the result of
> induction.
> > In so many cases, what is needed is that the ontological model
> be formative
> > in the context of a real situation, now; ie have a pragmatic
> dimension.
> >
> >
> >
> > -----Original Message-----
> > From: Harry Halpin [mailto:hhalpin@ibiblio.org
> <mailto:hhalpin@ibiblio.org>]
> > Sent: Monday, April 03, 2006 7:59 AM
> > To: Paul S Prueitt
> > Cc: 'Danny Ayers'; 'adasal'; semantic-web@w3.org
> <mailto:semantic-web@w3.org>; timbl+speaking@w3.org
> <mailto:timbl+speaking@w3.org>;
> > colette.maloney@cec.eu.int <mailto:colette.maloney@cec.eu.int>
> > Subject: Re: [ontac-forum] Re: Semantic Layers (Was
> Interpretation of RDF
> > reification)
> >
> > I'm tempted to suggest that this conversation be moved to
> > semantic-web-philosophy@w3.org
> <mailto:semantic-web-philosophy@w3.org>. I'm also tempted to take
> a line from Pat
> > Hayes's argument against TimBL on whether or not a URI addresses a
> > single unambiguous thing, and just say "Look, you're not wrong,
> you're
> > just insane" as regards people who are complaining about the Web
> > lacking semiotics/pragmatics/the perfect design/a better syntax :)
> >
> > The entire point of the Web is that people have can create
> different
> > ontologies, which represent not necessarily agreeing points of view.
> > People can and will use different levels of abstractions and want to
> > talk about different things in differing manners, even using
> different
> > sorts of syntax. Despite this, by giving them the same formal
> semantics
> > and one naming system (URIs), they can actually use (owl:import) and
> > talk about each other.
> >
> > As soon as anyone says "I invented the *One Perfect Ontology*,
> and it
> > even includes very subjective things like *pragmatics* and
> *semiotics*,
> > so if everyone should use my one ontology and all our problems
> go away"
> > - well, I'd have to say that's a bad and naive idea. Assuming
> there is
> > "The One Big Ontology" out there we can all use endorses a naive
> > logical positivism (a sort of blatantly wrong reading of the
> Tracatus)
> > and this sort of thinking has been ditched by both philosophers and
> > psychologists (as well as most ordinary people). There is a giant
> > well-documented literature in philosophy and psychology that (no
> > surprise) shows our perceptions and abstractions are
> situation-specific
> > - I would recommend the work of Andy Clark for easy-to-read
> > introductions. I would say that the same applies to the "Look at My
> > Great Design" argument that Sowa was advocating earlier.
> >
> > So, yes, just implement a standard upper ontology of pragmatics and
> > semiotics (in KIF, OWL, whatever) and then e-mail the listserv
> when it
> > actually does something useful using a real-life use-case
> instead of
> > complaining that the Semantic Web doesn't map directly onto it, and
> > people will be pleased. You may even win the RDF.net prize!
> >
> > But even then it will never solve everyone's KR problems, and the
> > entire point of the Semantic Web isn't to endorse "One Big Ontology
> > based on Bunge" but to allow people to create their own small
> ontologies
> > in a decentralized manner. And that may be a good idea.
> >
> > Paul S Prueitt wrote:
> >
> >> You suggest in
> >>
> >> " The RDF/OWL view doesn't really make a distinction between
> Upper Level
> >> Ontologies and Domain Ontologies, but it has been demonstrated
> that ULOs
> >>
> > can
> >
> >> be expressed in RDF/OW"
> >>
> >> That there exist upper level ontology that meets all
> requirement imagined
> >>
> > in
> >
> >> Semantic Web language and that it has been demonstrated that
> this upper
> >> level ontology can be expressed in OWL?
> >>
> >> Is this what you are suggesting?
> >>
> >>
> >>
> >> -----Original Message-----
> >> From: Danny Ayers [mailto: danny.ayers@gmail.com
> <mailto:danny.ayers@gmail.com>]
> >> Sent: Monday, April 03, 2006 5:50 AM
> >> To: ONTAC-WG General Discussion
> >> Cc: adasal; John F. Sowa; semantic-web@w3.org
> <mailto:semantic-web@w3.org>; Paul S Prueitt;
> >> brian.macklin@cec.eu.int <mailto:brian.macklin@cec.eu.int>;
> timbl+speaking@w3.org <mailto:timbl+speaking@w3.org>;
> >>
> > colette.maloney@cec.eu.int <mailto:colette.maloney@cec.eu.int>
> >
> >> Subject: Re: [ontac-forum] Re: Semantic Layers (Was
> Interpretation of RDF
> >> reification)
> >>
> >> On 4/3/06, Azamat < abdoul@cytanet.com.cy
> <mailto:abdoul@cytanet.com.cy>> wrote:
> >>
> >> Simply put,
> >>
> >>
> >>> we must understand which web (or architectural pillars) most
> fits the
> >>> matter, the formal semantic web ( i.e., the syntactic web,
> known as the SW
> >>> layer cake) or the real semantic web, something like this version:
> >>>
> >>> <Real Semantic Web> ::= <Ontological Framework> < Logical
> Framework>
> >>> <Semiotics> <the Web>
> >>> <Ontological Framework> ::= <UFO> <Upper Level Ontologies> <Domain
> >>> Ontologies> <EOL>
> >>>
> >>> <Logical Framework> ::= <FMF> | < ... > <EOL>
> >>>
> >>> <Semiotics> ::= <Pragmatics> <Semantics> <Syntax> <EOL>
> >>> <Pragmatics> ::= <Users> <Web Agents> <Intentions> <Actions>
> >>>
> >>>
> >> <Communication>
> >>
> >>
> >>> < Proof, Trust> | <Truth> <EOL>
> >>>
> >>> <Semantics> ::= <Signs, Natural Language Expressions>
> <Meanings> <EOL>
> >>>
> >>> <Syntax> ::= <Rules> <OWL Ontology> <RDF Schema> <RDF M&S> < RDF>
> >>>
> >>>
> >> <XML/SGML>
> >>
> >>
> >>> <Namespaces> <EOL>
> >>> <the Web> ::= <Resources, state, representation,
> identification, URI,
> >>> Unicode> <Interaction, sofware agents, hypertext links,
> protocols, HTTP>
> >>> <data Formats, HTML, XHTML> <EOL>
> >>>
> >>>
> >> I'm neither a philosopher nor logician, so forgive me if sounds
> naive:
> >> how does the above "grammar" conflict with what (if I understand
> >> correctly) you are calling the "syntactic web" - i.e. the
> Semantic Web
> >> of the W3C initiative?
> >>
> >> Ok, there are certainly differences, like here:
> >>
> >>
> >>
> >>> <Ontological Framework> ::= <UFO> <Upper Level Ontologies> <Domain
> >>> Ontologies> <EOL>
> >>>
> >>>
> >> The RDF/OWL view doesn't really make a distinction between
> Upper Level
> >> Ontologies and Domain Ontologies, but it has been demonstrated that
> >> ULOs can be expressed in RDF/OWL.
> >>
> >> ...here:
> >>
> >>
> >>
> >>> <Pragmatics> ::= <Users> <Web Agents> <Intentions> <Actions>
> >>>
> >>>
> >> <Communication>
> >>
> >>
> >>> < Proof, Trust> | <Truth> <EOL>
> >>>
> >>>
> >> and here:
> >>
> >>
> >>
> >>> <the Web> ::= <Resources, state, representation,
> identification, URI,
> >>> Unicode> <Interaction, sofware agents, hypertext links,
> protocols, HTTP>
> >>> <data Formats, HTML, XHTML> <EOL>
> >>>
> >>>
> >> - only half of each of these are explicit in the layer cake,
> the rest
> >> (I would suggest) being implicit parts of the system, e.g. the
> >> Semantic Web being an extension of the current Web, the current Web
> >> includes HTTP hence the SW includes HTTP. Both feature Users,
> Agents
> >> etc.
> >>
> >> So it looks to me like your "real semantic web" is the same as the
> >> W3C's Semantic Web, but for a few undocumented features in the
> latter.
> >> Where's the problem?
> >>
> >> Cheers,
> >> Danny.
> >>
> >>
> >>
> >> --
> >>
> >> http://dannyayers.com
> >>
> >>
> >>
> >>
> >>
> >>
> >
> >
> >
>
>
> --
> -harry
>
> Harry Halpin, University of Edinburgh
> http://www.ibiblio.org/hhalpin 6B522426
>
>
--
-harry
Harry Halpin, University of Edinburgh
http://www.ibiblio.org/hhalpin 6B522426
Received on Wednesday, 5 April 2006 00:19:37 UTC