W3C home > Mailing lists > Public > public-owl-dev@w3.org > October to December 2006

Re: OWL "Sydney Syntax", structured english

From: Adrian Walker <adriandwalker@gmail.com>
Date: Thu, 30 Nov 2006 14:41:18 -0500
Message-ID: <1e89d6a40611301141q6b070b8al1c9ca7128ebdb938@mail.gmail.com>
To: "John McClure" <jmcclure@hypergrove.com>, public-owl-dev@w3.org
John --

Thanks for your note, and congratulations on the design of the
hypergrove.com web site.

It seems to me that there are two, partially overlapping Semantic Web
visions.

The first concerns the kind of work you are doing, which I believe is mainly
about bringing order and accessibility to text documents.

The second concerns what I take to be Timbl's other vision -- a web-wide
database of RDF triples.  So the data is structured (as triples), rather
than textual.

I guess there is some commercial success in parsing text documents to
extract (meta)data.  However, automatically parsing English knowledge and
converting it to logic for reasoning seems to be a much harder task, at
least at the industrial strength level.

Our  Internet Business Logic work, with its lightweight approach to English
knowledge input, is mainly directed to reasoning over structured RDF and
other data, although there are  some examples such as [1] that reason about
documents.

So, the aspect of RDF that we mainly care about is that it allows you in
principle to freely mix and match structured data from different sources on
the web.  There's actually more to it than that, though [2].

The example [3] is the closest we have got to document exchange so far.  As
you may see, the ontological aspects are in executable English rules**,
rather than in OWL.

There are also some small OWL-related examples, such as [4].

Perhaps one place where our respective approaches begin to overlap is this.
Wwe do a form of information retrieval to try to tie an English question
that  a  user has typed in to the concepts that are currently loaded into
the system.

Best regards,  -- Adrian

**  As previously mentioned, the rules are open vocabulary, open syntax.

[1]  www.reengineeringllc.com/demo_agents/RDFQueryLangComparison1.agent

[2]  www.semantic-conference.com/program/sessions/S2.html

[3]  www.reengineeringllc.com/demo_agents/SemanticResolution1.agent

[4]  www.reengineeringllc.com/demo_agents/{OwlTest1 OwlResearchOnt
FeaReferenceModelOntology2}



Internet Business Logic (R)
Executable open vocabulary English
Online at www.reengineeringllc.com
                                Shared use is free
Adrian Walker
Reengineering
Phone: USA 860 830 2085




On 11/30/06, John McClure <jmcclure@hypergrove.com> wrote:
>
>  *Adrian,*
> **
> *I am curious about this fascinating approach -- may I ask *
> **
> *(1) if there is no ontology (your words: "no vocabulary or grammar
> construction"), why do you care about the RDF which depends completely on
> class and property definitions?  If your response is that "the approach
> creates classes and properties as a consequence of the text analysis" then
> is the resultant ontology ever stored? or re-used? or shared?*
> **
> *(2) is "document exchange" out-of-scope (inapplicable) for this approach,
> since there appears to be no contractual reference ontology between
> publisher and consumer?*
> **
> *Thanks much for your reply,*
> *John*
>
> -----Original Message-----
> *From:* public-owl-dev-request@w3.org [mailto:
> public-owl-dev-request@w3.org]*On Behalf Of *Adrian Walker
> *Sent:* Thursday, November 30, 2006 9:00 AM
> *To:* Pat Hayes; public-owl-dev@w3.org
> *Subject:* Re: OWL "Sydney Syntax", structured english
>
> Pat --
>
> You wrote...
>
> There have been several proposals for English-like
> syntaxes for logic, see for example John Sowa's 'structured English'.
> Again, one can make these look quite convincing by a deft choice of
> basic vocabulary, but they always become incomprehensible when one
> uses a slightly divergent one. The problem is that when it reads
> *almost* like English, any non-English constructions - nouns in place
> of verbs, the wrong preposition, etc., - become very intrusive and
> awkward. Some object-oriented programing notations claim similar
> transparency, and there have been proposals for English-y syntaxes
> for KRep notations, such as various frame-based systems which allow
> things like (Every Person who owns a donkey beats the donkey of
> self). I confess to not having citations ready for this, but such
> systems were developed at U. Texas, for example.
>
> Yes, there are many proposals to try to model enough of ordinary English
> usage to make writing and running knowledge easier than with formal
> notations.  The underlying idea in all of these is to parse with a grammar,
> translate automatically to some form of logic, and to execute that.  There
> are brave folks who also attempt the reverse translation, from logic to
> English.
>
> As has been pointed out many times, this approach does not seem work
> outside of natural language research projects.   If it did work, it would
> surely by now be a huge commercial success. It appears to encounter several
> roadblocks, including the ones you mention.  The fact that English is a
> moving target  does not help.
>
> There is a different approach.   The approach is lightweight,  and seeks
> to go around the deep NL research problems involved, rather than tackling
> them head on.  Roughly speaking, the approach is to assign an open
> vocabulary, open syntax string to each predicate symbol in the underlying
> logic.  If a predicate is n-ary, the corresponding string has n place
> holders (or variables) such as "some-person" or "that-time".    There's more
> to it than that, but that's the basic idea.
>
> This allows one to label predicates with strings such as
>
>     so far as is known at this-time there is no evidence to suggest that
> this-person is a terrorist
>
> (Actually the approach starts with the string, and invents an arbitrary
> corresponding predicate say,  p33(x,y), for computation)
>
> This  lightweight approach means that there is no dictionary or grammar
> construction -- at least in the usual 'structured English' sense.   It also
> means that one can use jargon, government acronyms, 'google' as a verb, and
> so on.  Of course, this violates all sorts of expectations about how one
> should compute using English syntax and  semantics.   And it's of zero
> interest to NL researchers, rightly so.
>
> But, if one is willing to accept the trade off involved, it actually
> seems to be useful!
>
> As you may know, this is the approach taken for the author- and
> user-interface of the Internet Business Logic system [1].  The system is
> online, and shared us is free, so folks can check for themselves that they
> can write this kind of English to a browser, and then run it.
>
> BTW, my PHD thesis subject was Chomsky grammars, and like many other folks
> I have banged my head dutifully against the 'structured English' wall.
> Great research topic.  Very hard to make it work at industrial strength.
>
> With apologies to Kendal,     -- Adrian
>
> [1] Internet Business Logic (R)
> Executable open vocabulary English
> Online at www.reengineeringllc.com
>                                 Shared use is free
> Adrian Walker
> Reengineering
> Phone: USA 860 830 2085
>
>
Received on Thursday, 30 November 2006 19:41:36 GMT

This archive was generated by hypermail 2.3.1 : Wednesday, 27 March 2013 09:32:53 GMT