Re: [xml-dev] Word of the day: upconversion

[Liam Quin <liam@w3.org>]

On Fri, Aug 21, 2009 at 08:59:36AM -0400, Costello, Roger L. wrote:
> Recently I read an article [1] by Michael Kay and learned a fabulous word:
>     upconversion

I think I first heard this when we used Omnimark at SoftQuad --
it supported upconversions, downconversions, and... I can't remember
what it called (calls?) xml to xml (SGML to SGML) conversions.

> Consider the spectrum from random text to well-designed XML:
> 
> <------------------------------------------------->
>   random                                       XML
>                upconversion --------->
> 
> 
> Suppose we place prose somewhere between random text and XML:
> <------------------------------------------------->
>   random         prose                         XML
>                upconversion --------->
> 
> Where would you place XPath?

I wouldn't, it's orthogonal.  In your example (quoted below)
you are using XPath to express a business rule, but from
the conversion point of view, only the difference between the
prose and the XML is significant.

> For example, consider these three ways of expressing a business rule:
>     Level 1 managers may sign off on purchase requests 
>     that do not exceed $10K.
> 
>     not(purchase-request[number(cost) gt 10000])
> 
>     <Request id="purchase">
>         <signoff manager="level1" LE="10000" />
>     </Datatype>

A more usual conversion might produce,
    <rule>Level 1 managers may sign off on purchase requests
    that do not exceed $10K.</rule>

This could then be processed into
    <rule><staff level="1" role="manager"Level 1 managers</staff>
    may <action>sign off</action> on <item>purchase request</item>s
    that do not exceed <amount currency="USD">10000</amount>.</rule>

To get from this to recognising the "do not exceed" and to
understanding the use of "that" to mean "which" requires either
natural language processing, or a very carefully controlled vocabulary
in the input; usually I've found it's a case of recognising patterns
and then dealing with exceptions by hand.

Note that your XPath expression does not require that the person
be a manager, and does not handle the implied case that any staff
member can sign off on a purchase order below $10 000.
It may be a correct expression of a business rule, but it is possibly
not a good result of an upconversion of the text.

Liam



-- 
Liam Quin, W3C XML Activity Lead, http://www.w3.org/People/Quin/
http://www.holoweb.net/~liam/ * http://www.fromoldbooks.org/