> -----Original Message-----
> From: Z39.50 Next-Generation Initiative [mailto:[log in to unmask]]On Behalf Of
> Eliot Christian
> Sent: Thursday, July 22, 2004 10:06 AM
> To: [log in to unmask]
> Subject: Re: Demo business case for search interoperability
>
>
> At 10:29 AM 7/22/2004, Peter Noerr wrote:
> >[...]
> > > Of course, if the end goal is to achieve something like "result
> > > clusters within facet", then this final step is Really Difficult
> > > (ergo, a key selling point of the standard search service).
> > >
> >This is confusing two aspects. You are actually contemplating a
> "federated
> >search" ("metasearch") vs. a "sequential search". The outline
> has nothing to
> >do with use of standards or otherwise,
>
> Sorry, I left out a piece of the logic chain. One of the key
> inhibitors for products like Vivisimo is the need to retrieve
> results that have well-known structure other than just rendering
> mark-up. So, Vivisimo could metasearch three distinct Z39.50
> sources and retrieve results in a chosen schema. Then, the
> clustering engine could create separate clusters for "author",
> "publication date", "subject", etc., in addition to whatever
> clustering may be desirable against the "body of text".
> Absent the Z39.50 standard, all they can do is to pick apart
> whatever the engineers assume about the HTML returned by the
> various servers.

This was a part of the work flow I assumed to be there. However, the
metasearch engines do this normalization from a wide variety of sources.
Standards like z39.50 help a lot (but are not a complete answer due to
incomplete and unequal implementation) however they do not provide complete
source coverage in almost any universe of targets so the metasearch engines
do the individualized mapping to provide data in a consistent form the likes
of Vivisimo can digest.



>