Peter,
Verity Search 97 Information Server does not provide support for DC "out
of the box".
However, one of the features of S97IS is that you can add whatever
fields you want, through a configuration file called "style.ufl" (style
files configure various aspects of the Verity "collection", the UFL is
the user-defined fields list).
The drawback of Verity is that multiple instances of a DC field (eg:
DC.Title = "My Document", DC.Title = "Untitled 1 - My Document") must be
combined in some form before Verity indexes the document. If you don't
combine them, Verity will only record the last one it saw.
eg:
DC.Title = "A Seminal Paper Investigating the Productivity Gains of
'Open Sourcing' Software"
DC.Title = "Use The Source, Luke"
Would have to be combined into the form:
DC.Title: "A Seminal Paper ... Software"; "Use the Source, Luke"
Verity would then index the DC.Title field as containing
"A Seminal Paper ... Software"; "Use the Source, Luke"
And you could search on that content using/building a query of the form:
DC.Title <CONTAINS> "Open Source Software"
Verity S97IS also ranks matches against field searches as a percentage
of the number of items a document matches. If you have one item (the
example search above), then any documents that contain "Open Source
Software" in the title will get 100%. If you were to search for DC.Title
<CONTAINS> "Open Source Software" <AND> DC.Title <CONTAINS> "Use the
Source", you would get a range of documents with 100% (both terms), 50%
(one term) or 0% (no terms).
If you combine this with a full text search, you will get different
results, because the full text search relevance is calculated using some
arcane method that I don't understand.
Now that I've got that out of the way and actually read your question
again... here's how you could implement schemes in Verity Search 97
Information Server (not that I've actually done most of these):
Given a tag of the form
<META NAME="DC.Date" SCHEME="WDTxxx" CONTENT="1999-06-15">
You would need to munge that to something like
DC.Date: 6/15/1999
(where DC.Date is defined as being a date field)
OR
DC.Date = 199906150000
(where DC.Date is defined as being a number)
OR
DC.Date = (WDTxxx="1999-06-15")
(where DC.Date is defined as being a string)
Note that storing a date or number as a string negates any attempts to
search within a range. Other schemes could be encoded similarly:
DC.Author = (VCARD="name:Satrapa, Alex;email:[log in to unmask]")
(blatantly displaying my lack of knowledge of the vcard format)
DC.Field = (SCHEME="content");(SCHEME="content");(SCHEME="content")
Then you can look for
DC.Author <CONTAINS> <WILDCARD>(VCARD=*name:Satrapa*)
I haven't flexed Verity's <WILDCARD> operator muscle yet, so I'll have a
further look and see how well this one will work. Perl programmers will
realise immediately that the question must be asked of how "greedy" the
* operator is in that wildcard pattern.
Anyhow, I digress. If you have further questions, please contact me at
this e-mail address.
Regards
Alex
[log in to unmask] wrote:
>
> Does anyone on the list have any official information from
>
> Microsoft, specifically regarding Site Server support for Dublin Core (e.g.
> I'm told Site Server won't accept periods in names)?
> Verity, specifically regarding support for DC schemes?
> Pete Winn
> Knowledge Management InformationTechnologies
> Information Services International - Los Angeles, CA
> Email: [log in to unmask]
> Voice: (973)691-3853 (MTO ext. 3853)
> Fax: (310)446-1626
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|