Unifying our cultural memory: Could electronic environments bridge the historical accidents that fragment cultural collections?
in Information Landscapes for a Learning Society, Networking and the Future of Libraries 3, 1998. and presentation at UK Office of Library Networking Conference, July 1998.
David Bearman and Jennifer Trant, Partners, Archives & Museum Informatics, USA
(Section 4)
There are several different kinds of problems encountered in attempting to translate between schemas. Examples are necessary to explore each of these points, since the principles here seem very abstract. Again, the Getty/CIMI examination or users' questions is helpful.12
- The simplest problem
is that the value assigned to an attribute being sought is not in
the vocabulary of the user. Karol Schmiegle recounted the following
observations from the Winterthur Museum "We may record that the
object is made of leaded glass, but the person seeing this material
may wonder why, if the object is made out of lead I, it can be so
transparent."13
- The attribute may
be present in both the user's model and the documentation system,
but no value has been assigned by the cataloguer. "Do you have any
film of the last V-1 to fall on London?" requires a knowledge of
both the nature of a "V1" and a complex temporal inference.
- The attributes being
sought by the user, though present in his worldview, may not be
present in the worldview of the repository. "What images of women
are there that are created by gay men?" implies a level of detail
in both artist's biography and image subject analysis that may not
be documented.
- The attribute may be present, but it may belong to a different entity. "What items were found in the Sutton Hoo horde?" may require searching documentation of the excavation, in order to find it by its vernacular name, and then finding objects that were excavated at a particular, scientifically identified site.
But many other questions inferred a knowledge far beyond that of one information resource.
- "Do you have any miniatures
depicting the 'churning of the milky ocean'?" requires a detailed
knowledge of Indian Iconography to know where, in what object genres
and which contexts this might be shown.
- "When were fish knives
first used?" requires determining the questioners meaning of "fish
knife," understanding the concept of "first use" and the purpose
of the questioner/ It could be answered equally well by a Breugel
image of fishing that shows a knife, a Victorian fish service, and
a fish slice, which is a serving tool.
- "What is the essence
of minimalist Art?" may not be answerable, and certainly wasn't
asked to generate a list of titles linked to examples of minimalist
art. An article on the subject might be more appropriate.
- Knowing the purpose of the user is essential to determining an appropriate answer. "When was the Mona Lisa on exhibit at the National Gallery of Art?" takes on a new meaning when it is asked by the Andy Warhol Museum.
But often, this survey revealed that real questions reflect assumptions and misconceptions, and that providing an appropriate response often involved a number of stages of refinement, and disambiguation. A favorite question from a teacher, 'Do you have any pictures of interiors designed by Mr. Rococo, suitable for framing?' revealed just how bundled. Once the confusion of "Mr. Rococo" was dealt with, the attributes of what is really being sought here are very complex. Queries about the availability of a reproduction, genre of art, implied subject matter, audience interest level - would return little from traditional collections documentation systems.
It is easy to dismiss "Mr. Rococo" but we can't ignore other more accurate queries for "paintings by modern French artists suitable for copying for my tenth grade world history class study of inter-war France". These create equally complex problems. Copyright status of the works of modern European painters is highly unlikely to be recorded anywhere because it is known as a consequence of the general provisions of copyright laws. Subject matter is seldom recorded in art museums, and here the user is seeking a subject with an implied relevance to history beyond being art. Genre terms such as painting are correctly applied only to the painting itself, so a photograph or poster of the work (which is in fact what the user is seeking) would, in art repositories be described as a photograph and not as a painting. Finally, the nationality of the artist is an attribute of the artist and not the painting, and is, in addition, not as straightforward as generally imagined, since the nationality of an artist at the time that they painted a particular painting may not be that which is generally recorded in biographical databases (the nationality at time of either birth or death). Information recorded as an attribute of the painting, which looks like the nationality of the painter, may be the place the work was painted.
With correct knowledge of the values being sought by the user in a normalized model of the domain, the system can begin to seek resources that might satisfy the inquiry. Note, however, in our example that the operative term is what the teacher wants to do with the result. It could have been a publisher seeking an illustration or a dealer seeking to buy the work of art itself - the limitations of a successful inquiry are that it must yield a result suited for the stated purpose of use. Models must recognize that goal of the query may not be an object or artifact, but a process or procedure that will guide a future action, or provide context for a past one.14 The verbs are as important as the nouns.
If we imagine our school teacher looking for "modern French paintings" again, the system could reasonably assist this user at the discovery phase of research by correctly modeling the query itself. Is the user really interested in an artist whose nationality is French, regardless of their location during the inter-war years, or more interested in art created in France, regardless of the nationality of the artist, or perhaps (and this is most likely given the target audience and pedagogic purpose) of artists whose subject matter was France. By explicitly displaying the mapping that the system presumed, and perhaps prompting to show (from previous discourse frames recorded in a query database) that other possible models exist, the system could disambiguate the initial query, and help the user to refine her question into one that could produce the desired result.15
NEXT: Mechanisms: metadata declarations, common models, and shared vocabularies