At 09:34 AM 8/31/2004, you wrote:
> > -----Original Message-----
> > From: Eliot Christian [mailto:[log in to unmask]]
> > Sent: Tuesday, August 31, 2004 5:43 AM
> > To: [log in to unmask]
> > Subject: Re: Abstract concepts as search indexes
> >
> > How about something like this:
> >
> > Developers of context sets should be aware that searchable content will
> > be mapped to search indexes and then results will be mapped to a record
> > schema separately. This can lead to confusion in the case where an index
> > has the same name as a record schema element. For example, a server may
> > map the "dc.title" index to several MARC fields in a database yet map
> > the "dc.title" record schema element to only MARC field 245$a. Although
> > both mappings are sensible individually, the searcher would see result
> > records that may not have the searched words in the "dc.title" element.
> > The general recommendation, therefore, is that an index name be the
> > same as a record schema element name only if both are to be mapped to
> > the same content.
>
>Nope. How about:
>
>Creators of search interfaces need to be aware that the schema they use to
>display the contents of the target database may provide an incomplete view
>of the data. This may result in the user being confused about why the
>record was returned. Creators should either warn their users that this
>might happen or retrieve the complete record.
In an application that searches multiple formats, there is no notion
of a particular "complete record". Again, the concern here is using
a named abstract concept for search and using that same name in the
retrieved result. Such usage is valid for elements that typically map
the same content in search and retrieval (e.g., "postalCode"), but it
is a problem for common concepts like "title" that are may not be mapped
to the same content in search as in retrieval.
|