Would generating browsable forms from parsed components mean that the
analyzable information related to a person's name that could be tagged in
MADS would be limited to elements that are desired for a fixed browsable
string? For example, there are many names for which a date becomes known
which are already authorized without a date. In such cases, would there be
a restriction on putting dates in a "dates" part of the name data cluster?
Would there be a need for a separate "dates not in heading" dates field? It
might be simpler to have an unparsed field for "authorized heading" for
browsing use, separate from the parsed set of name parts needed for
searching and other kinds of manipulation. Of course, this only matters if
the headings in the file are supposed to be consistent with the forms in
external authorizing files (e.g., NAF). If the MADS headings only represent
entities in a particular file of content object records that they're linked
to, updating the heading form with new data ceases to be an issue.
As for searching in the latter scenario, I'm assuming that any content
object file that has been linked to a MADS file would route searches of
terms defined in MADS first to the MADS records, and retrieve only them.
Searching with a common term could recall lots of MADS records, but
assuming one could search "joyce" for example as a "family name," I doubt
the problem would be worse than a comparable browse search in most cases,
and the possibilities for search refinement and sorting alternatives with
parsed data elements from the MADS record would be greater. Once the
desired entity or concept has been selected from among the returned MADS
records, then the MADS record ID is used to retrieve the corresponding
content objects. This points up the importance for this scenario of being
able to tell from the MADS record whether one has found the name or work or
topical concept which one is seeking before moving on to the content
objects. The disadvantage would be having to take at least two steps to get
to the content objects. The advantage would be more precise content object
retrieval, and not having to create unique, structured heading forms to
differentiate entities. Unique record IDs and unique, searchable,
expressive MADS records would suffice, and would probably be much easier to
create than MARC AACR and LCSH authorities.
So again, my basic question is--what needs is MADS intended to serve? The
MARC Authorities format as it's been applied is focused more on browsable
heading and reference forms, and less on clearly identifying and enabling
complex searching of entities and concepts; but the latter may be more
important in the MODS environment where MADS is expected to operate, and
the former less important. The greater the variety of uses envisioned for
it, the more complex the structural requirements of the schema will be to
allow for different and possibly conflicting applications.
Stephen
At 10:14 AM 6/11/2004, you wrote:
>On Fri, 2004-06-11 at 09:01, Stephen Hearn wrote:
> > If we don't care about browsable lists
> > anymore, then we don't need carefully structured heading strings.
>
>I'd say that we don't need to carry those "browse" headings as fields in
>our records, but we should be able to generate them (easily, I'd say)
>from marked-up forms. So
> <name>
> <forenames>John James</forenames>
> <familyName>Smith</familyName>
> </name>
>
>Easily becomes:
> Smith, John James
>in a browse list.
>
>BTW, ONIX has a variety of name forms defined that seem to meet some of
>these needs:
>
>PersonName - the name in normal order: James J. Johnson III
>PersonNameInverted - the name in sort order: Johnson, James J., III
>
>Then it has a 7 part name sequence:
>
>TitlesBeforeNames - i.e. Sir
>NamesBeforeKey - before the key name, i.e. any forenames: James J. (note
>that this assumes that the group of forenames will always be treated as
>a single unit -- there is no attempt to break this down further)
>PrefixToKey - i.e. van
>KeyNames - those names usually used to sort by: Garcia Marquez, or
>Madonna (This is a nifty solution for the one-word names)
>SuffixToKey - i.e. Jr., III
>LettersAfterNames - i.e. Ph.D.
>
>If you have the coded 7-part name, you do not need either for the first
>two (person name and person name inverted) because you can generate
>them. Then again, if you receive names as whole strings and don't have
>the means to accurately parse them, you can use the fields that carry
>full names. They are less flexible, but they should always result in a
>user-friendly display.
>
> > I can
> > imaging a file of MADS records which would generate responses to a search
> > of discrete elements associated with an author. Multiple MADS records might
> > have identical "headings" or display forms, but contain enough additional
> > information in the MADS record to distinguish each author. By linking bib
> > record name elements to the correct MADS records, one could have retrieval
> > by individual authors without having individuated author headings.
>
>It sounds like you are suggesting that we use info in the record to
>distinguish authors, rather than forcing the headings to be unique. This
>could facilitate dialogues like:
>
>"Do you want James Joyce, who wrote Ulysses, or are you looking for
>Joyce James, the cookbook author?"
>
>The trick here would be finding a very brief display of the data from
>the record that could work when the user has typed in "Joyce" and
>retrieves a number (in many library catalogs, that means thousands) of
>entries.
>
>kc
>
>
>--
>-------------------------------------
>Karen Coyle
>Digital Library Specialist
>http://www.kcoyle.net
>Ph: 510-540-7596 Fax: 510-848-3913
>--------------------------------------
****************************************************
Stephen Hearn
Authority Control Coordinator
Projects, Data and Sciences Team Leader
University of Minnesota
160 Wilson Library Voice: 612-625-2328
309 19th Avenue South Fax: 612-625-3428
Minneapolis, MN 55455 E-mail: [log in to unmask]
|