“Data or Document? Migration of Descriptive Metadata for Medieval and Renaissance Manuscripts Between Data-Centric and Document-Centric Models: A Case Study”

Elizabeth J. Shaw Aziza Technology Associates ejshaw@azizatech.com

This paper will discuss a case study of the dual direction migration of metadata describing medieval and renaissance manuscript collections between the Digital Scriptorium database schema and an XML DTD developed to encode extant manuscript descriptions. This discussion highlights the challenges and tensions that exist in trying to merge data that is captured in different data models, roughly categorized as data-centric and document-centric. The observations garnered from this study may have broader implications for other efforts to meld distinct descriptive models for unitary searching and resource discovery.

CONTEXT

The author was hired by Digital Scriptorium to:

1)provide XML —> HTML XSLT transformations of an electronic version of the Huntington Library “Guide to Medieval and Renaissance Manuscripts” which has been encoded in the TEI Medieval Manuscripts Description Work Group (TEI MMSS) draft DTD ,
2) migrate data from the existing Digital Scriptorium database to the the draft DTD and
3) migrate the TEI-MMSS encoded “Guide” to the Digital Scriptorium’s existing database schema.

lIn the context of this work, many interesting challenges have come to light. The Digital Scriptorium database schema and the TEI MMSS draft DTD are closely allied since some of the same people worked on both data models. However, significant differences exist in nomenclature and structuring of the data, both because the TEI MMSS DTD is an extension of TEI and because the purpose of the TEI MMSS DTD is to allow for the capture of extant descriptive information. The database schema, created prior to the DTD is a more constrained data model and uses slightly different elements and nomenclature for its basis.

DESCRIPTIVE TRADITIONS

Although the library community has largely settled on a particular record-based model for descriptive information (the MARC record and AACR2) about printed works, serials and selected media, many communities of practice that have need of description to facilitate resource discovery and scholarship have no such common standard. In many cases, a narrative form of description has evolved in an attempt to capture the particular descriptive needs of the curators and users of the materials. Descriptive practice in disciplines such as manuscript collection or archives has varied greatly across nations, institutions and within individual institutions over time. This evolution has largely been dependent on the interests, energies and capabilities of individual curators. Furthermore, descriptive practice varies greatly between communities of practice because of the particular characteristics of the disciplines that they support. Both the elements of description fundamental to supporting work and the nature of their expression may be unique. That these descriptive needs vary greatly has become more obvious as efforts to develop single metadata standards or standards that easily map across disciplines have had limited success. Nonetheless, these communities often face similar problems as they attempt to migrate their formerly print based description to an electronic form that can be shared across institutions. Each community must revalidate what aspect of description is important to the work of the community, 2) define a common way of expressing information about those important things and 3) develop a model that accurately captures that information. The work must be done in an environment with no uniform extant practice and differing philosophies about the nature, scope and role of description in the discipline. This paper will focus on two approaches to capturing significant descriptive data within the same discipline and examine the challenges in merging the two.

DATA-CENTRIC VS. DOCUMENT CENTRIC DATA MODELS

The numerous schemes for descriptive metadata that have emerged since the advent of the web vary greatly in their scope, complexity and underlying assumptions about the nature of descriptive practice. Ronald Bourret’s article XML and Databases and his associated materials provide an interesting framework from which to view these different approaches to capturing descriptive metadata. The emerging practices might be categorized into two distinct models:

Data-Centric models: These record-like models capture discrete data elements in a structure that is well represented in a flat or relational model (ie. Dublin Core, MARC). Often these flat or relational representations atomize data at the most discrete levels at which one might manipulate it. Some require or enforce the normalization of the data, institute data typing and constrain the syntactical expression in which it is represented. The Digital Scriptorium database schema follows this record-like model, though without some of the rigor of syntax that is required in a MARC record.
Document-Centric models: These models capture more narrative or discursive, often complex, hierarchical descriptions. Often, they allow considerable flexibility in what data elements are required, the form in which data is expressed, and rarely (especially prior to the advent of XML Schema) utilize data typing to constrain the expression of information.

These document-centric models may further be categorized as those that provide a representation of an idealized model of descriptive practice and those that represent extant descriptive practice. Those that attempt to capture an electronic edition of an extant description are often the least rigorous in their requirements for data elements and normalization because of the need to accommodate heterogenous practice. The tension between representing an idealized model of a particular document structure—either to facilitate machine processing of the data or to enforce practice in a community—and providing a model that captures existing representations has been discussed extensively by authors such as Piez and in efforts to develop data models such as the TEI dictionaries DTD

CAPTURING EXTANT DESCRIPTION

Although document-like descriptions from various institutions and time frames seem similar in format they often contain distinctly different elements of description in distinct orders and with variant nomenclature. This has presented challenges to the developers of DTDs and schemas that attempt to capture descriptive metadata in these communities of practice. DTDs such as Encoded Archival Description (EAD) and the TEI MMSS draft DTD are explicitly designed to accommodate a variety of extant descriptive practice rather than constrain practice to the schema developers’ prescribed notions of practice. The extant documents that are encoded using these models often differ greatly from their record-like cousins. Narrative and complex, they utilize indirect reference, inference from context (the description of a manuscript written in French may never explicitly state the language in which it is written, assuming that the reader can discern from the quoted text imbedded in the description), inference from relationships to siblings or parent elements, and assumptions about relationships of the parts of the description that are not explicit. Although a human reader can infer these relationships, the lack of explicitness makes it more difficult to capture characteristics of the described object in machine processable ways. To capture this information explicitly requires modifying the extant description. Data-centric models tend to capture this information in explicit ways from the start.

MIGRATING DATA

Many practitioners and users of existing descriptive material find that even when description is reworked by knowledgeable humans in order to import it to a data-centric metadata representation, the records fail to capture significant information that is embedded in the narrative of these extant descriptions. In many cases, the original cataloger has specialized knowledge of either the collection being described or of the field of study. References to other significant or related work, connections between objects in the collections are often lost in their transformation to data-centric descriptions Tensions between models for capturing descriptive metadata are often amplified at the point at which migration occurs. Automated efforts to migrate encoded extant description have met with mixed success. Chris Prom’s migration of EAD encoded descriptions to Dublin Core RDF records for importation into a cultural heritage OAI database points to some of the problems inherent in automated migration of extant data. Lundberg discusses the implications for information retrieval of data encoded in XML.

THE DIGITAL SCRIPTORIUM: CASE STUDY

This paper reports on the outcomes of the effort to automate the migration of data captured in two distinct models and the resulting implications for resource discovery. The paper describes:

the differing data models,
the process by which the migration occurred,
the challenges faced both in migration from relational database to XML and in migration from XML to a relational database model,
the development of a relational database model that attempts to capture both the XML encoded manuscript descriptions as well as the data in the existing database
the trade-offs in these various representations and
the implications for information display and resource discovery.

By highlighting points of ambiguity, strengths and weaknesses of the different approaches to metadata capture and challenges in merging the data into the same repository, it is hoped that this study will inform the work of others who face similar challenges in their efforts to provide descriptive metadata to their user communities.

REFERENCES

Ronald Bourret. XML and Databases. : ,

. Digital Scriptorum. : ,

. Digital Scriptorium Database Schema. : ,

. Encoded Archival Description. : ,

Sigfrid Lundberg. “Excursions along the border between metadata for resource discovery and for resource description.” . : .

Wendell Piez. “Beyond the ‘descriptive vs. procedural’ distinction.” Extreme Markup Languages. Montreal, Canada: , 2001. 197-214.

Christopher Prom. “Does EAD play Well with Other Metadata Standards?: Searching and Retrieving EAD using the OAI Protocols.” Journal of Archival Organization. 2003. 1: .

. TEI dictionaries DTD. : ,

. TEI Medieval Manuscripts Description Work Group (TEI MMSS) draft DTD. : ,