Re: Attribute Architecture and completeness question

On Thu, May 08, 2003 at 09:33:23AM -0400, Shuh Barbara wrote:
> Alan,
> You can find the latest Bib-2 Attribute Set documents at 
> 	http://www.nlc-bnc.ca/iso/z3950/bib-2_home.htm
> 
> On that page, there is a link to Mapping Bath Profile Searches to Z39.50
> Attribute Architecture Sets
> 	http://www.nlc-bnc.ca/iso/z3950/bib-2_in_bath_1_0.pdf
> This should help answer your questions on how Bath Profile searches are
> mapped to the new AA.

Thanks. Reading through this I have a new question on the philosophy
behind some of the attribute lists. I am trying to work out how to
implement a AA binding in our system.

Quick background - we usually index access points as words and as a
complete string. Right truncation of complete titles at character
and word boundaries use the complete string index. So I am trying
to work out what attribute lists map to a word index vs a complete
value index. With Bib-1 its easy - its the completness attribute.

With the Bib-2 bindings for Bath there no longer appears to be a
single attribute I can use to determine which index to use.
So I started mapping out the cases in a big table. All word based
queries specify

    (5) Expansion/interpretation [Bib-2] = (2) all of these words

All of the complete index queries do not. But if I ignored Bath for
the moment, in the general AA sense is this intended? That is,
are searches by default anticipated to be on the complete value of
fields and that if you want to do a word based query you have to
use the Bib-2 attribute set? This does not feel right to me. I
assume I am missing something.

In particular, 'Title search - exact match (2.01. %.A.1.6)' seems
strange to me.

    (1) Access [Cross-domain] = (1) title
    (8) Comparison [Utility] = (3) equal
    (5) Expansion/interpretation [Utility] = (1) adjacent words
    (9) Format/structure [Bib-2] = (1) set of words

If the default is not implicity that you are querying on complete
values for access points, what in the above indicates that the
complete value of the field is being searched? Does 'adjacent words'
imply that the words are also left and right anchored? Naively this
would have been the query I would have thought would have been looking
for the set of words as a *subset* of the title. Eg: 'Roman Empire'
would match 'The Fall of the Roman Empire'. But the Bath profile
says this will not match - the title must be exactly 'Roman Empire'.

Note 'set of words' does not imply matching on the complete value
as 'title search - keyword' also uses 'set of words'. In fact the
only differency is the (5) Expansion/interpretation value is
[Bib-2] (2) all of these words.

I would have thought perhaps it should have 'left anchored' and
'right anchored' added - the anchoring then implies it doing
a complete value match instead of a word level match.

> Where did you find the broken links?
> There is a reference to the correct Bib-2 Attribute Set documents in the
> site index on the Z39.50 MA.
> 	http://lcweb.loc.gov/z3950/agency/contents.html
> and from the page on the Attribute Architecture
> 	http://lcweb.loc.gov/z3950/agency/attrarch/attrarch.html
> 
> Barb

Either pixies fixed it in the night, or I was in the wrong window
and looking at a cached version on my laptop. Now only the music
link in http://lcweb.loc.gov/z3950/agency/attrarch/attrarch.html
is broken. (It has two URLs in it separated by a space.)

Thanks again,

Alan

Received on Thursday, 8 May 2003 19:23:59 UTC