2. Methodology
2. Methodology | |
The KSU Museum houses approximately twenty thousand items of costume and decorative arts from the eighteenth century to the present day. More than half of this collection has been photographed and documented. Additionally, over five thousand volumes of library material in the KSU library system provide supplementary material to support the Museum's exhibitions.
The sample for testing metadata treatment was drawn from two special collections of the KSU Museum: (a) about eighty American costumes identified for a project entitled "What Americans Wore, 1820-1920", and (b) some typical items from the Manchu robes collection. In addition, twelve non-costume items including some furniture, decorative arts, bedding, and porcelain items were also included. The KSU Museum stores all fashion items in its warehouse. These items are usually documented with accession records or exhibition labels. Some have archival documents such as a donor's letter or a record of an item's exhibition history. The prototype for the "What Americans Wore, 1820-1920" project was designed to include up to four views (front, back, profile, and detail) to represent each fashion object. The three image types, thumbnail, reference, and archival (per request only), each has different tonal depth, format, and spatial resolution. (The prototype is at: http://newmedia.kent.edu/waw/, please choose 1860 link.)
To conduct the metadata study, accession records for 90 objects and their accompanying documents and images were analyzed, and 42 items were selected for creating cataloging records. Many of the costumes contain more than one piece. For example, a female's costume may include a bodice, a skirt, and possibly a belt; a male's costume may include a frock coat, a waistcoat, and breeches. In the process of preparing cataloging records, the costume was treated as the object being cataloged, with necessary description notes for the components or accessories. When necessary, analytic records were also produced. Appendix B shows an example of an 1860 day dress with bonnet and shawl that belonged to Ms. Marie Eleanor Bente. A record was created for this object in which the dress was the main piece (see in Appendix C). Two analytical records for the bonnet and shawl were also created.
Although most of the study items have been photographed and many slides have been digitized, the original items -- the costumes, not the images of the costumes -- were used to prepare the records. The objects exist only as single objects and are not reproductions. This fact differentiates this project from similar ones based on reproductions or digital images alone.
It was assumed in developing this online fashion collection that the cataloging format should follow the principles and standards of object description already established by information professionals. Standards not only guide how a collection is described and how individual values are normalized, but also enhance interaction and interoperability with the collections of similar institutions (Marshall, 1998). When the project started, there were several established metadata standards. They were primarily created for describing document or document-like objects, not for three-dimensional museum objects. "The museum community has not been able to agree, even within a discipline, on what constitutes a core record for the description of museum objects" (Taylor, 1999, p. 97). It was essential for the project team to determine whether any of the existing standards would satisfy descriptive needs for fashion objects. Among the key questions to be addressed were the following: Would it be possible to adopt and enhance existing metadata standards to accommodate three-dimensional artifacts? In this regard, did a need exist to develop a new metadata format or sets of elements? Would other kinds of three-dimensional objects also require individual metadata sets?
Literally dozens of metadata schemes were available at the time as this project progressed, ranging from popular generalized metadata formats to specialized metadata formats. These formats included:
Each of these formats was constructed from an understanding of specific domains, information resource needs, and unique requirements for describing document and document-like objects. Though not particularly designed for describing non-document-like objects, each had some applicability.
In the current study, three metadata formats were examined: AACR2 in use with the United States MAchine-Readable Cataloging (USMARC) format, the Visual Resources Association Core Categories for Visual Resources (VRA Core), and Dublin Core Categories (Dublin Core). Two local considerations contributed to the selection of these formats: compatibility and simplicity. In regard to compatibility, the project leaders wanted to ensure interoperability with related collections. Considering USMARC as one option was based on a plan to merge fashion object records with the KSU library online catalog, KentLINK, and also to provide compatibility with the Library of Congress American Memory project, which was the original impetus for the digitized fashion collection. VRA Core was first brought to the project team's attention by the Ohio Library and Information Network (OhioLINK), of which KSU is a member. In early 1998, OhioLINK initiated a multimedia database project and recommended that the VRA Core should be used as a record structure for multimedia content in the art and architecture domain (OhioLINK, 1998, p.2).
In regard to simplicity, both the American Memory digital library project and the OhioLINK multimedia database project require that collections be easily accessible by a variety of users, without requiring special browsers or additional plug-ins. Another factor is that the KSU Museum staff had no experience with formal cataloging except in efforts to maintain an in-house dBase database for registration purposes. The Dublin Core was attractive because it enables any author of an electronic publishing product to create minimum-level cataloging records or surrogates. VRA Core's orientation to visual resources and art works and its simplicity and flexibility also suggested that this format could be a good match for this project.
The format testing stage consisted of four steps. First, fifteen costumes were selected and divided into three groups. For each group of five items, one of the three formats was applied as the primary format to create basic records. The records were then converted into the remaining two formats. As a result, each of the fifteen items had records in three formats. Second, the author, a research assistant, and the museum curators analyzed the records to assess the strengths and weaknesses of each format when applied to a specific costume item. Richness of content and structural presentation in a record as well as productivity and quality of cataloging were major considerations. The museum staff favored the VRA Core. They also suggested additional elements. With the supplementary elements, the modified VRA Core format was selected to be applied to the entire prototype collection. (However, the decision did not represent a final judgment about which format is acceptable or not acceptable for describing fashion objects. See discussions in later sections.) Third, a modified data dictionary based on the VRA Core Categories was developed and a template for generating records was created. Twenty-seven additional items were cataloged to further test the template. Fourth, in anticipation that other union collections (such as the KSU Library's KentLINK and OhioLINK's Digital Media Center) might want to include KSU's fashion collection in their databases, two format-mapping tables (from the extended VRA Core format to USMARC and to Dublin Core formats) were developed for the purpose of conversion.
Image description was considered for future exploration, since fashion apparel represented on the Web by a set of images are "reproductions" and therefore they become objects to be cataloged/described. The decision to separate or to integrate original work descriptions with the visual document descriptions remains undecided. Therefore, only descriptions of original work received consideration at this stage.