Re: The author of a query is (not) most likely the user

Hi Bijan --

Thanks for your further comments on authoring, reasoning, and 
explanations.  I have taken some time to read your paper -- 
http://www.mindswap.org/papers/debugging-jws.pdf -- hence the delay in this 
reply.

First let me say that I'm impressed that your paper contains a useability 
study.  That's rare and valuable.

My take on your paper is that we addressing complementary parts of the 
authoring/debugging problem.

In SWOOP/PELLET + debug, you focus on internal inconsistencies in an 
ontology, and you explicitly set aside any attempt to help the author with 
the "subtler" problems of
inferences that (a) happened and should not have , and (b) that should have 
happened but did not.

In our 	INTERNET BUSINESS LOGIC system, we are mainly concerned with 
exactly those "subtler" problems.  We are also trying to be helpful to 
inexperienced authors, whereas your focus is on folks with at least 9 
months' experience of OWL.

So, for example, our system will explain in hypertexted English why an 
unintended inference happened.  It will similarly explain why an intended 
inference did not happen.

To be picky, I'm concerned that while the SWOOP-PELLET GUI looks nice, in 
Figure 3 in your paper you had to add a hand-written English explanation of 
what the figure means.  That could be taken as an argument for expressing 
the knowledge in the ontology in lightweight English in the first place, so 
that the machine can produce an English explanation.  What do you think?

Thanks for taking the time to look at our system (at 
reengineeringllc.com).  You wrote that you found the example

"""customer some-cust-number is on the some-type plan effective 
some-plan-start-date
that-plan-start-date is less than or equal some-current-date
not : customer that-cust-number has switched plans between 
that-plan-start-date and that-current-date
""""
to be less than natural English.  You are right.  For lightweight *open* 
vocabulary English like this, there *is* a trade off.  It's sort of like 
following a mountain path with a cliff to the left and a canyon to the 
right.  The cliff is Full Natural Language Processing, that no-one seems to 
know how to do robustly and with feasible maintainability.  The canyon 
is,  er, techie notations like raw OWL.  Clearly, a GUI like SWOOP is 
another way of finding such a "cliff path" between AI-complete problems on 
the left and machine-oriented human-unreadable stuff on the right.

You wrote also:

"My current belief is that it's hopeless to expect (most) end-users to be 
able to handle any sort of direct authoring of rules. They could maybe do 
wizards and forms, depending. But any \"just write it down\" seems like a 
hopeless non-starter."

Well, yes, you need a certain amount of skill level.  But less than 9 
months'  worth (:-)   A business analyst writing a spreadsheet is perhaps a 
useful example to compare to?

To see why it may be useful to close up the gap between business folks and 
techies, consider the following, from

     http://www.financetech.com/showArticle.jhtml?articleID=163701206

"Now, data truly is ubiquitous, touching all corners of an enterprise
and becoming the lifeblood of an organization. Not only must data
managers be savvy technologists capable of keeping that information
pure and pumping it through the corporate veins, they also must be
able to speak the language of business. They must be astute business
advisers and diplomats who understand not only the data, but also the
business lines they serve - including equities, fixed income and
derivatives - as well as the problems business line managers face."

So, how about OWL-LE ?   (OWL plus lightweight English)    (:-)

			Cheers,    -- Adrian

Received on Tuesday, 12 July 2005 20:49:22 UTC