Re: The imprecision of Z39.50

On Mon, Jul 07, 2003 at 12:12:14PM +0100, Robert Sanderson wrote:
> So we need a format/structure of 'string', or for that to be recognised as 
> the default if no format/structure is given (which seems silly given the 
> existence of the above No Expansion, we should be as explicit as possible)

I agree 100%. In the separate proposal I just mailed I went further to
suggest Any/All/Adj words is not a clean indiciation of format/structure
but rather instructions of how to deal with multiple terms in the query
string and that there should be a single attribute value to say 'words'. 
Otherwise you potentially have to scan on Title + AnyOfTheseWords etc
which feels semantically incorrect.

Note: Bib-2 defines some funky format/structure values for dealing with
names which I think is a good example of usage of the attribute type.
I don't fully understand it, but it seems to be saying things like
'last-name, first-name' vs 'first-name last-name'. These attributes
are clearly the describing the format/structure of the term with no
built-in query operator (unlike AnyofTheseWords). This is directly
in line with the description of format/structure given in the AA overview.

> You could supply more than one expansion attribute?
> Otherwise you also couldn't do case insensitive matching with stemming, 
> and so forth.

You can repeat expansion attributes, so no problem there. The AA
defines for class 1 which attribute types can have repeating values,
and what the repeating values mean.


There are lots of ways to approach the problem. It is possible to tack
it on the side. But I would rather get it clean, especially since I don't
think the AA is really being used much (at all?) yet. I don't see how
to fit And/Or/Prox (All/Any/Adj) into any of the existing attribute types
(at least with a straight face! ;-).

Alan

Received on Tuesday, 8 July 2003 00:19:05 UTC