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

Re: OWL "Sydney Syntax", structured english

From: Bijan Parsia <bparsia@cs.man.ac.uk>
Date: Mon, 4 Dec 2006 04:07:59 +0000
Message-Id: <A0D931D6-4740-42A6-BBFD-A3C05428BB15@cs.man.ac.uk>
Cc: "Kaarel Kaljurand" <kaljurand@gmail.com>, public-owl-dev@w3.org
To: Adrian Walker <adriandwalker@gmail.com>

On Dec 3, 2006, at 8:01 PM, Adrian Walker wrote:

Sometime, I don't know when, since people in this thread  
have...interesting...quoting discipline, Kaarel Kaljurand wrote:

>> I played with the Internet Business Logic. The idea is elegant.  
>> But is
>> it useful?  Have you done studies showing that people prefer IBL  
>> to, say,  Prolog.

The term "studies" in the above text seem to refer, most naturally,  
to controlled experiments, though, plausibly, it could refer to  
longitudinal or case studies, or perhaps user surveys.

> Thanks for saying the idea is elegant.  Hopefully, studies such as  
> [1,2,3] are building evidence that it is also useful.

The "studies" cited in this subsequent text do not seem to fall into  
any of the paradigms I mentioned. One could, perhaps, stretch, and  
call them case studies, though, as such, I do not find them  
especially illuminating. Perhaps it's best to call them "worked  
examples". The first (<www.reengineeringllc.com/ 
Oil_Industry_Supply_Chain_by_Kowalski_and_Walker.pdf>) is perhaps  
more of a system description or white paper with a worked example  

Furthermore, it does not seem to me that there is an active research  
program, at least. I'm pretty sure I read the first article some time  
ago, and the second examples have the strings "Version 20041215" and  
"version 20041029" which strongly suggests that they are from 2004 (I  
also recall seeing these or similar from at least that time frame).

I recall an email thread twixt me and Adrian wherein he explicitly  
said that he didn't have formal studies:

"""The problem can get worse with rule systems, if we stick to the  
techie notations that us techies know and love.  Or it can be  
mitigated, if we make sure that the person writing the rules has to  
document what they mean at the real world business level.  If this is  
done by including lightweight English in the rules themselves, the  
rules and the documentation cannot get out of step.  We can also get  
English explanations of the reasoning.

We don't yet have HCI  studies to say whether this is a good  
approach. As you said elsewhere in your posting, such studies are  
hard to do, and sometimes inconclusive.  However, a system** that  
supports the approach is live, online.  So, one can get an idea of  
the 'value proposition' (ouch!) by viewing and running the RDF and  
other examples provided.  One can also use a browser to write and run  
one's own examples.  Non-commercial use of the system is free.""""

(Note, that this was specifically in the context of the need for  
explanations, and explanations in NL.)

I don't mind advocacy, and often proof is in the pudding and the  
proving is in the eating. I personally am happy to document my own  
idiosyncratic and native speaker reactions. But can we *please* not  
pollute the discussion space with such misclassification and  
hyperbole? In either direction? (E.g., Pat's skepticism about NL/CNL  
techniques also carries no water with me and I *am also* suspicious,  
by inclination, of NL/CNL techniques!).

I know four things in this area:
	1) our pilot study was surprisingly promising for understanding (so,  
search results; verification by domain experts, etc.)
	2) someone reported using Swoop's NLP view for verification and even  
correction by domain experts (that last part I don't understand quite  
as we don't provide a parser)
	3) I know that there are ontology building teams that make use of  
stilted NL toward CNL as a core part of their methodology with crappy  
tool support
	4) Some users have expressed some happiness at Manchester syntax,  
though there have been reports that it's still "too logicy"

I thought the last was mentioned by the presenter of this paper:

(Which is well worth reading, folks! And it includes a study study :))

But I cannot find this in the paper itself. So I'm unsure about 4.

(The paper is a good reminder that, well, our tools suck. I say this  
as a toolbuilder! I can see places in the paper where I can predict  
how swoop would do, and it's not pretty :) Sigh. I mostly know, I  
think, how to fix a lot of it, but it's also true that Swoop was not  
designed for "normal" users. Though I don't know how "normal" the  
users are if they are comfortable in an Eclipse based environment :)).

Received on Monday, 4 December 2006 04:08:23 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:58:14 UTC