Re: D7 - Which languages are better than OWL?
Dear Philippe,
Your wiki is probably an interesting one - unfortunately, I have not
time enough to look at this in detail. I am in fact heavily engaged in a
(concrete) application of NKRL (Narrative Knowledge Representation
Language) in the "Defence" domain, and I have to sent to Springer the
final manuscript of my NKRL book by the end of December. Therefore,
please find below only some quick, incomplete and disorganized comments.
- The main problem of the W3C languages is their lack of expressiveness:
they are "binary" languages, and a lot of important applications asks
instead for the use of "n-ary" languages. All this is perfectly well
known, but it is also apparently "politically incorrect" to evoke this
truth too often. Documents on the n-ary extensions of the W3C languages
in the style of "www.w3.org/TR/2006/NOTE-swbp-n-aryRelations-20060412/"
do not tackle really the problem: the majority of the examples in this
document do not concern the "n-ary problem" at all but only the way of
adding properties to the "standard" binary relations, and the unique
real "n-ary" example dealt with is managed using a rehash of old
semantic networks solutions of the '70. This lack of expressiveness of
the W3C languages also hampers the development of some real good ideas
that have blossomed in a W3C context, like Semantic Web Services.
- From the "expressiveness" point of view, moreover, the idea of
establishing the central version of OWL on Description Logics (DLs) has
been probably a very unfortunate idea. Already dismissed at the
beginning of the nineties because of their uselessness from a practical
point of view, DLs have been resurrected in a SW context in name of a
distorted interpretation of the "sound and clean semantics" exigency -
you can easily imagine how much the average programmer will care in
general about computational tractability. DLs are limited, ugly, verbose
and, moreover, difficult for this same average programmer to deal with:
this easily explain why OWL is, until now, a real flop from a
concrete/commercial point of view. See in this context, among other
things, the recent survey on the Semantic Web Vision published in the
IEEE Intelligent Systems issue of September/October of this year. From
an "applicability" point of view, RDF is probably more useful (see also
"commercial" products like Oracle 11g RDF database, or the
RDF-compatibility of recent Web 2.0 products like GroupMe!) once, of
course, its limitations have been clearly acknowledged.
- With respect then to "rules" - how do you think to build up some sort
of concrete application without rules? - the situation is not really
better (and probably worst). Based on a combination of OWL Lite and OWL
DL, SWRL is already "naturally" limited - and, ironically, it cannot
support OWL Full and RDF/RDFS. Moreover, being "semi-decidable", DLs
gurus strongly suggested to make use only of its decidable subset
DL-Safe-SWRL, with the results that the SWRL variables can only be bound
to known individuals in a knowledge base! In practice, to make use of
rules in a W3C context, it is still better to use Jess or Algernon via,
e.g., a plug-in of the Protege-OWL plug-in. Note that Jess is nothing
than a re-implementation in Java terms of CLIPS, OPS5 etc. which, once
again, bring us back to the seventies and the expert systems era...
To conclude (quickly and provisionally): give up with OWL and
OWL-based applications, there are better and more "expressive" tools
around, from Conceptual Graphs to NKRL etc. But - as I repeat regularly
from the beginning of the SUO endeavour - don't hope of being able to
make use of these already existing building blocks to set up a better
alternative to the W3C language without regular project(s), money to
hire people, to meet each other "face to face", to build up significant
applications etc. Informal discussion are surely useful, but they will
never build up a system.
Season's greetings,
Gian Piero ZARRI
University Paris4/Sorbonne and CityPassenger
France
Philippe Martin wrote:
>Hi all,
>
>Here is the announcement to start the discussion on
>D7. What languages are better than OWL for developing ontologies,
> and what are their pros and cons?
>
>I think there are at least two tasks w.r.t. this discussion:
>- arguing for or against the need for certain features such
> as expressivity, readability, conciseness, normalizing-effect
> in a "general" language for developing ontologies,
>- categorizing languages with respect to these features or the
> elements that contributes to these features.
>Any other task needed?
>
>For the discussions to have an organised and precise result,
>- I created a wiki page with examples of semantically structured
> discussions and tool comparison tables for the above topic,
>- whenever you want to make a contribution, you are invited to
> extend the wiki page and post to the mailing list the updates
> you made (a simple-copy paste of the addition you made is
> sifficient),
>- if updating the structures on the wiki page is too difficult,
> I invite you to at least re-use (and quote) terms and statements
> of the wiki page in your informal contributions on this
> mailing list; other persons (me, at least) will then be able
> to update the wiki page adequately.
>Any other structure needed?
>
>This wiki page is currently at
>http://krl.wiki-site.com/index.php/Main_Page
>(in case you encounter problems accessing this page, a static copy
> is at http://www.phmartin.info/suo/).
>Please have a look, and please forgive (and correct) the possible
>mis-categorisations due to the lack of time.
>This wiki page will be moved to the "IEEE wiki for the SUO" when it
>becomes available.
>Any other suggestions?
>
>Philippe
>
>
>