Robert Sanderson wrote:

>First draft is available at:
>   http://srw.cheshire3.org/contextSets/rel/
>
>Comments, as always, are sought :)
>
>

Having the possibility to specify a ranking/ordering mechanism in a
query is a nice feature ... buuuut not enough in the type of federated
search engines we try to build (and either not enough for peer-to-peer
merging of relevances).

What is really needed for federated search and merging of hit sets is:

1) an by the appropriate relevance ranking algorithm ordered response
set from
    any of the SRW servers - that's what you are adressing

2) an indication of the actual relevance value of every record in the
hit set (this
    needs a new SOAP tag in  SRW)

3) an indication of the possible range of  relevance values for each server
    implementation - I can not meaningful merge  ranked hit sets of two
servers if
    the one assigns ranking values [0, 10.000] and the other [-inf,
+inf]. So the servers have to tell me in what  range relevance values
will be.
    Being a purist and a matematician, I vote for the default rank range
[0,1]

Obivously, ther must ne made place for 2) and 3) in the  SRW protocol.

Given these three pieces of information, I have the possibility to build
multitarget searching with really cool and correct merging of hit sets
according to the specified ranking algorithms.

Marc Cromme, Index Data


>Rob
>
>--
>      ,'/:.          Dr Robert Sanderson ([log in to unmask])
>    ,'-/::::.        http://www.o-r-g.org/~azaroth/
>  ,'--/::(@)::.      Special Collections and Archives, extension 3142
>,'---/::::::::::.    University of Liverpool
>____/:::::::::::::.
>I L L U M I N A T I  L5R Shop: http://www.cardsnotwords.com/
>
>
>