Re: [semanticweb] Something that fundamentally confuses/bothers me about the Semantic Web
So you want someone to prove you wrong.... I can't do that quite yet, but
there is a group that I have been working with that is driving steadily
towards a good solution. It's come out of the aerospace industry, but is
gathering steam in internet architecture groups, government and ebusiness
The Universal Data Element Framework http://www.udef.org is being designed
as a bridge solution, and we believe that we will create a framework that
will be easy to maintain by domain subject matter experts and easy to
include into the huge variety of dictionaries that represent the semantic
meaning, context and structure (RDF/OWL, schema, DTD, and other formats).
This looks like a centralized authority at first glance, but will end up
being an open standard based registry that can be replicated, like the DNS
and IP routing tables, across the net. There will also be a repository
where domain formats for eBusiness, Semantic Web or other apps can make
their UDEF tagged ontologies or taxonomies available to users via several
access settings (public, subscriber only and private).
There is a recent article at Network World:
You're right on the exact point that will dramatically lengthen the time it
takes to stand up the Semantic Web (and automated eBusiness for that
matter): Each ontology or dictionary is fairly specific to the subject area
that it deals with. So either everyone has to use the same ontology or
build compatibility into all their ontologies, taxonomies, vocabularies,
etc. The problem is huge, and the solution in the corporate world so far
has been to create millions of point-to-point mappings between, for
example, the warehouse management system, the parts ordering system and the
supply chain company systems that feed the warehouses and factories.
Currently I'm standing up the General Motors partner integration framework,
where we will use ebXML based web services to communicate with all the
different systems across the entire automotive supply chain. Most of the
auto industry is trying to standardize on OAGIS for XML business document
format, but we buy from electronics (RosettaNet), petroleum (PIDX),
chemical (CIDX) and others.
The UDEF is a tree / branch structure with 18 top level object words (such
as entity, document, assett, etc) and 16 top level property words (such as
Date, Amount, Identifier, Quantity, etc.) with qualifiers that are
assembled into a semi-intelligent ID. For example,
purchase.order.document_identifier is expressed as d.t.2_6. Placing this as
an attribute in a schema or RDF enables the semantic equivalency between
formats. You can view the entire trees in graphical form (HTML version of a
powerpoint) at http://www.udef.org/specdoc/UDEFv1pt03-July-2003.htm. There
are XML versions of these trees that will be used in the registry system
(link to the use cases is below) and as a communications format to other
We have made some good progress but the hard work is ahead, as the registry
stands up, and we begin to solicit input from domain area experts to add
the qualifiers that reflect their industries... most of the qualifiers have
been added by people in aerospace, defense, government and automotive. We
will need input via a collaborative web environment (planned for in the use
cases) of many experts. Please volunteer by sending me email or sending
mail to the udef.builders mailing list.
Dave Hollander, one of the inventors of XML, and CTO at Contivo, on the UDEF:
"Contivo actively supports the Universal Data Element Framework (UDEF).
Semantic interoperability is a long term goal. Reaching that goal will
require the ability to understand vocabularies quickly and easily.
The ability to deploy UDEF concepts within vocabularies represented in a
variety of languages makes UDEF unique, interesting, and potentially a
corner-stone of semantic interoperability. Contivo is evaluating how to
implement the UDEF to reach this potential."
There is a proof of concept you can use two modified XML docs (OAGIS 7.1
and xCBL purchase orders) to produce a semantic equivalency report with, at
my web site: http://www.sanghainteractive.com/whitepapers.html at bottom,
labeled "UDEF Compare Report".
There are a couple of OASIS groups that are interested in working with the
UDEF, and we will most likely begin a UDEF Technical Committee within OASIS
You can view the use cases that we will be building the registry on this
year, at http://www.udef.org/specdoc/master_list.html
I would encourage anyone interested to join the udef.builders mail list, by
sending email to udef.builders-subscribe@...
At 05:38 PM 7/4/2004, adamatlas wrote:
>I hope someone can explain this. I was just as excited about the Semantic~~~~~~~~~~~~~
>Web as everyone
>else for a while, but then I realized something which discouraged me.
>The Semantic Web, it seems, is being billed as a system that allows the
>machine-readable data in interoperable, extensible, decentralized formats.
>But this seems
>to create a paradox, or at least a technical contradiction. It simply
>can't be all three. It's a
>classic "choose two" problem. If we want RDF-based formats to be
>can't be extensible, because inevitably some mutually incompatible
>vocabularies will arise, or they can't be decentralized since some
>authority will have to
>maintain this interoperability. If we want RDF-based formats to be
>extensible, then they
>can't be interoperable for the same reason.
>I witness lots of semantic wheels being reinvented constantly. eBiquity
>has their Person
>which duplicates much of the
>functionality of FOAF. There are several other FOAF-alikes that I've seen,
>vocabularies, ones which are not specifically about people but involve
>them, include their
>own "person" classes. How's that for interoperability? After all, the
>parser doesn't know
>that they're the same. In fact, if it's a good, well-formed RDF/XML
>parser, it explicitly
>thinks that they're different. That's why we have namespaces, isn't it? To
>similarly-named elements from different namespaces have different
>functions! But I think
>we can agree that an eBiquity person:Person is very similar in function to
>and that they should ideally be able to be used interchangeably the same
>It seems that the only ways to solve these overlaps for practical use are:
> 1) Hardcode these similarities in function into parsers. (BAD!
> Hardcoding is a last resort,
>and it'll be outdated in a heartbeat.) Or...
> 2) Create new ontologies for the express purpose of bridging these
> similar but
>incompatible formats, by describing the relations between the similar
>via relevant RDFS/OWL statements. (BAD! Who's in charge of doing this?
>want these formats to interact? Bad idea, too tedious and it negates the
>purpose of a
>standardized structure like RDF. The W3C? Bad idea, the last thing they
>need is to have the
>responsibility of constantly bridging new vocabularies. The authors of the
>themselves? Nonsense: if they wanted their vocabularies to be compatible,
>have made them like that in the first place.) Or...
> 3) Politely ask the vocabulary authors to add compatibility. (Better,
> sadly, but still BAD. If
>we're trying to create a web of pure, machine-readable data, it's simply a
>bad idea to rely
>on everyone voluntarily making their stuff interoperate with other
>people's stuff. Besides,
>as the number of vocabularies continues to grow, this will be a very
>tedious and inefficient
>Sadly, the current RDF-based system seems destined to become not a web of
>readable data, but a mishmosh of lots of mutually incompatible formats all
>happen to be based on RDF. Imagine our current World Wide Web but based on
>different XHTML-like formats that people can create and add to on whims.
>Do you expect
>it to work? We can't count on a "natural selection"-type process to weed
>out the ones that
>aren't as effective. Looking at
>[<http://www.schemaweb.info],>http://www.schemaweb.info], we see lots of
>identical vocabularies with trivial compatibility-destroying differences.
>doesn't occur in technology when the competing standards don't have any
>disadvantages over one another.
>... Or at least that's what it seems like to me. Believe me, I'd like
>nothing more than for
>someone to prove me wrong, considering how exciting the fundamental idea
>Semantic Web seems. So I'm open-minded about this, don't get me wrong, I'd
>just like to
>see if these concerns can be explained.
>Yahoo! Groups Sponsor
>Yahoo! Groups Links
> * To visit your group on the web, go to:
> * To unsubscribe from this group, send an email to:
> * Your use of Yahoo! Groups is subject to the
> <http://docs.yahoo.com/info/terms/>Yahoo! Terms of Service.
john c hardin
director / founder - sangha interactive
877.572.5691 - 313.737.1197 cell
"It is the framework which changes with each new technology and not just
the picture within the frame."
- In http://groups.yahoo.com/group/semanticweb/message/1664
john hardin a écrit :
> So you want someone to prove you wrong.... I can't do that quite yet,
Cannot this be told the other way around?
To prove Adam is wrong you would just have to prove
that *you* and other semantic web afficionados are right.
So, you cannot prove yet that you are right, can you?
> but there is a group that I have been working with that is driving steadilyEh!
> towards a good solution.
Then, how do you know you are "driving steadily towards a good solution" ?
Is it faith?
Or do you have some clues?
[skipping irrelevant technical mumbo jumbo]
> You're right on the exact point that will dramatically lengthen the time itIs it not that "Each ontology or dictionary is fairly specific",
> takes to stand up the Semantic Web (and automated eBusiness for that
> matter): Each ontology or dictionary is fairly specific to the subject area
> that it deals with. So either everyone has to use the same ontology or
> build compatibility into all their ontologies, taxonomies, vocabularies,
not only to subject area but *also* to specific usages within the subject area?
For the worse, it is even likely that each ontology would better be specific for each and every task or question within any subject area.
Clearly not achievable, where do you draw the boundary?
Or would you go drastic:
A single "world-wide" ontology encompassing *everything*, Orwell's nightmare!
OTOH, how to "build compatibility" between irreconciliable views from the *same* subject area like the alternative 3d 4d representation of space-time:
(from within http://www.arts.mcgill.ca/PROGRAMS/philo/faculty/mccall.htm)
And *don't* say that "this is only philosophy, we will never have nasty cases like that where business and industry stuff matters".
You *already* have such nastynesses and *lots* of them, you are just not aware!
>The problem is huge, and the solution in the corporate world so farYes, it works, based on trade-offs and expensive hand tuning.
> has been to create millions of point-to-point mappings between, for
> example, the warehouse management system, the parts ordering system and the
> supply chain company systems that feed the warehouses and factories.
How do you expect to dispense with the "hand tuning" given that the trade-offs *themselves* depend on each and every specific case at hand?
> Currently I'm standing up the General Motors partner integration framework,.../...
[more technical mumbo jumbo]
> We have made some good progress but the hard work is ahead,Yes, indeed! :o))
You will find out how painfull it is.
And this is just some kind of parts inventory with some lucky restrictions in the variety,
i.e. you may have a large number of different nuts and bolts but their total number
is (relatively) stable, new and old models *and* their properties don't pop up and disappear like mad at any unexpected moment.
You get some relief between the updates even if only a few days (or may be hours?).
> as the registry stands up, and we begin to solicit input from domain area experts to addHeading toward hand-crafting a monolithic "all encompassing" ontology.
> the qualifiers that reflect their industries... most of the qualifiers have
> been added by people in aerospace, defense, government and automotive. We
> will need input via a collaborative web environment (planned for in the use
> cases) of many experts.
> Please volunteer by sending me email or sending
> mail to the udef.builders mailing list.
Will crash and burn, sometime...
> Dave Hollander, one of the inventors of XML, and CTO at Contivo, on the UDEF:Long term sure, could you give a figure?
> "Contivo actively supports the Universal Data Element Framework (UDEF).
> Semantic interoperability is a long term goal.
> Reaching that goal willBefore that, it will require what "to understand vocabularies" *means* inside a computer!
> require the ability to understand vocabularies quickly and easily.
[still technical mumbo jumbo,
underlying problems not solved,
not even STATED!!!]
> "It is the framework which changes with each new technology andExcellent quote!
> not just the picture within the frame."
> Marshall McLuhan, 1955
Except today, the framework changes every hour.
Enjoy the ride.
Beware of the wall at the end of the road!
-- Jean-Luc Delatre
"In a way, math isn't the art of answering mathematical questions,
it is the art of asking the right questions" -- Gregory Chaitin
http://perso.club-internet.fr/jld/ -- GSM: +33 6 11 24 06 29