metadata for music: issues and directions
TRANSCRIPT
METADATA FOR MUSIC: Issues and DirectionsAuthor(s): Sherry L. VellucciSource: Fontes Artis Musicae, Vol. 46, No. 3/4 (July-December 1999), pp. 205-217Published by: International Association of Music Libraries, Archives, and Documentation Centres(IAML)Stable URL: http://www.jstor.org/stable/23509267 .
Accessed: 15/06/2014 22:07
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp
.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].
.
International Association of Music Libraries, Archives, and Documentation Centres (IAML) is collaboratingwith JSTOR to digitize, preserve and extend access to Fontes Artis Musicae.
http://www.jstor.org
This content downloaded from 185.2.32.121 on Sun, 15 Jun 2014 22:07:42 PMAll use subject to JSTOR Terms and Conditions
METADATA FOR MUSIC Issues and Directions
Sherry L. Vellucci (St. John's University, New York)
Metadata is data that describes the attributes of a resource, characterizes its relation
ships to other resources, supports the discovery, management, and use of a resource,
and exists in an electronic environment. Many different metadata sets exist, ranging
from general schemes to accommodate information in a wide variety of disciplines and formats, to highly specialized schemes designed for specific formats or subject communities. Common issues relating to all schemes include the lack of clear
descriptions for multiple "states" of a resource, conflict between discovery and
description functions, difficulty in describing aggregate resources, and lack of man dated standards for data content. The Dublin Core Element Set, fast becoming the
accepted metadata standard in the library community, must resolve these issues
before it can be the ideal metadata scheme to describe music resources. The Resource Description Framework (RDF), with its reference to linked authorities, may provide the optimal framework for implementing a variety of metadata to
describe music resources in a global network.
The term metadata has crept into the vocabulary of most librarians, music and
non-music, cataloguers and non-cataloguers. It has the power to enthrall
because it holds so much promise. Metadata will organize the Internet; it will
promote interdisciplinary research on a global scale; it will accommodate the
description of as yet unknown formats; it will provide unprecedented control
over electronic resources; and while doing all this, it will enhance the power and, therefore, the reputation of those who create it. Does the reality of meta
data live up to these promises? For now, at least, the answer is no. But move
ment in the metadata direction is fast indeed, and music librarians cannot afford
to be left behind in this brave new world of information organization. Unless we
take a leadership role in the development of music metadata, other less knowl
edgeable people will make the rules by which we must eventually play. Does this mean that all of the detail-laden MARC records will be converted
into simple Dublin Core metadata records? Probably not! But music librarians
will be creating and cataloguing more and more electronic resources, includ
ing digital facsimiles of music treatises, scanned images of music scores, sound files of archival recordings, and videos of staged productions. Along with these new digitized resources comes the ability to embed metadata with
in these digital objects, or link metadata descriptions directly to the files. The
205
This content downloaded from 185.2.32.121 on Sun, 15 Jun 2014 22:07:42 PMAll use subject to JSTOR Terms and Conditions
206 FONTES ARTIS MUSICAE 46/3-4
MPEG-41 and the developing MPEG-72 ISO standards allow metadata to be embedded within multi-media objects using a synchronized metadata track.
Already libraries are migrating to integrated library systems (ILS) that can contain and simultaneously search several metadata databases. Thus, music librarians will not only need to understand a variety of metadata schemes in order to enable data interchange within their own systems, but will be respon sible for creating metadata records in formats other than MARC. A firm under
standing of the important issues in metadata development will assist in this transition.
The Current State of Affairs
The most common definition of metadata is "data about data." This is a simple and accurate definition, but it needs a bit more elaboration. The term "meta data" began to appear frequently in the literature on database management systems (DBMS) in the 1980s, and was used to describe the information that documented the characteristics of information contained within databases.3 In this DBMS domain the computer was the setting for both the information
being described and the descriptive data, and, therefore, metadata operated within a totally electronic environment.
The parallel world of library cataloguing traditionally used the terms "bibli
ographic data" or "cataloguing data" for this type of descriptive information, even when describing computer files. But when cataloguers began to describe networked electronic resources using the exact same type of bibliographic data, things changed. Suddenly the MARC record became "metadata." What happened to cause this transformation? The answer is: convergence with the broader world of information organization. The methods of organizing resources from the rather separate domains of library science, computer sci
ence, and information science all converged in this networked environment, and, for better or worse, the term "metadata" became the commonly accepted term in all three disciplines. On the plus side, metadata provides a common term to use when communicating with other information organizers in the computer and information science world. It allows everyone to interact
equally on the continuum of information organization. This equality is impor tant for the development of bibliographic systems that access, transfer, and
manipulate metadata information across multiple databases in a way that pro vides meaningful information to the user. On the down side, the term "meta data" allows information organizers in the computer and information science domains to continue to ignore not only the efficient and effective process of
"cataloguing" that developed over the last century but also the very large body
1. International Organisation for Standardisation. "MPEG-4: Overview (Melbourne Version)." Web page, October 1999 [accessed 4 January 2000]. Available at http://www.cselt.stet.it/mpeg/ standards/mpeg-4/mpeg-4.htm.
2. International Organisation for Standardisation. "MPEG-7: Context and Objectives (Version 10)." Web page, October 1998 [accessed 4 January 2000], Available at http://www.cselt.stet.it/
mpeg/standards/mpeg-7/mpeg-7.htm. 3. John T. Phillips, "Metadata: Information About Electronic Records," ARMA Records
Management Quarterly 29, no. 4 (1995): 52-57.
This content downloaded from 185.2.32.121 on Sun, 15 Jun 2014 22:07:42 PMAll use subject to JSTOR Terms and Conditions
METADATA FOR MUSIC: ISSUES AND DIRECTIONS 207
of cataloguing literature that already exists on many of the so-called "metadata issues" under discussion today.
In order to place metadata within its proper context, the following expanded definition, which incorporates statements of functionality and environment, is used. Metadata, therefore, is data that:
• describes the attributes of a resource (name, title, publisher, dates, etc.)
• characterizes resource relationships (Is Based on, Is Version of, Contained in)
• supports resource discovery, management, and effective use (descriptive data, administrative data, terms and conditions data, intellectual property rights data, content ratings data, provenance data, structural data), and
• exists in an electronic environment.4
Although this definition reflects the metadata ideal, in reality most metada
ta schemes do not accomplish all of these functions equally well. Descriptive data come in a variety of levels and detail. Metadata schemes designed for use
by the general Internet-user public tend to concentrate on very basic data for
resource discovery and retrieval (e.g., HTML <META> Tags). libraries prefer more detailed descriptions that act as surrogates for the object (e.g., MARC
records). Only a few metadata sets express relationships adequately, although
implementation of the Resource Description Framework (RDF) as a formal
structural standard may change this in the future.5
Traditionally, metadata in library catalogues focused on the functions of
identification, description, and retrieval; but other metadata categories extend
into the areas of resource management and use. These include administrative
and structural data, terms and conditions data, content ratings data, and prove nance data. The extent to which any particular metadata set incorporates these
data categories varies depending on the needs of the community responsible for its development, and there are many such communities. Each group devel
oped its own metadata scheme to describe the resources most used by its
constituency. Some of these metadata schemes are general in nature—such
as the MARC format and the Dublin Core Metadata Element Set—and are
designed to accommodate information about electronic resources in a wide
variety of disciplines. Other metadata schemes are more specialized and apply to information in a specific format or within a specific discipline or domain.
What they all have in common is that they contain a set of defined data e
lements that describe the resource and help provide access to it in the elec
tronic environment. Beyond that, they all vary as to the number and content of
the data elements, and the standards used, if any, for that content.
4. Sherry L. Vellucci, "Metadata," m Annual Review of Information Science and Technology, ed.
by Martha E. Williams, 33 (1999): 187-222. (Medford, NJ: Information Today for the American
Society for Information Science, 1999.)
5. World Wide Web Consortium (W3C). "Resource Description Framework (RDF)." Web
page, December 1999 [accessed 5 January 2000]. Available at http://www.w3.org/RDF/ Over
view.html.
This content downloaded from 185.2.32.121 on Sun, 15 Jun 2014 22:07:42 PMAll use subject to JSTOR Terms and Conditions
208 FONTES ARTIS MUSICAE 46/3-4
Metadata Characteristics
All metadata schemes have three basic characteristics: semantics (or content),
syntax, and structure. The semantics can vary from many complex data elements
with content prescribed by standards and rules, to schemes that have as little as two or three elements with no control at all over that content. The syntax
may range from a highly complex coding system like the SGML to a basically undefined record structure like the original Dublin Core. Finally, metadata can be contained in a variety of architectural structures, including library cata
logues, Microsoft Access databases, or the recently formalized RDF standard. In addition to these three characteristics, there are three other aspects of
metadata that are the new buzzwords in digital information organization: flexi
bility, interoperability, and extensibility. Flexibility allows the metadata creator to include as much or as little detail as desired in the metadata record, within the overall constraints of the metadata scheme, with or without adherence to
specific cataloguing rules or authoritative lists.
Interoperability refers to the ability of information systems to interact in a useful way on a real-time basis over communications networks.6 For the sake of interoperability, many believe that metadata records should contain a basic core set of data elements common to all schemes. This would facilitate the
exchange and use of metadata in a variety of systems, and is the basic premise of the Dublin Core. Crosswalks offer a way to map the data elements from one metadata scheme to another and are a major part of the interoperability factor. Several crosswalks currently exist to map the MARC format to and from other
popular metadata schemes such as the Dublin Core.7 On the surface, cross walks appear to provide a simple means of exchanging data, for example the
DC.Title element maps to the 245 field in USMARC; however, few things in the world of metadata are as simple as they seem. No metadata crosswalks work
equally well in both directions. Mapping fifteen data elements into MARC may not seem difficult, but mapping a standard MARC record into fifteen simple Dublin Core elements presents a real challenge.
The final buzzword, extensibility, means that the scheme should allow the addition (i.e., extension) of data elements and data qualifiers to accommodate
specific user needs (domain specific). The danger with extensible data is a
lessening of interoperability, because the more that general metadata elements are qualified or extended on an individual or domain specific basis, the less
compatible or interoperable the metadata becomes with other schemes.
Metadata and the Music Community
While many subject-specific communities have developed their own metadata
schemes, the music community has not chosen that route as yet. Instead, most efforts have focused on evaluating the effectiveness of one general scheme,
6. Christine L. Borgman, "From Acting Locally to Thinking Globally: A Brief History of
Library Automation," The Library Quarterly 67, no. 3 (1997): 215-49.
7. Michael Day, "Metadata: Mapping between Metadata Formats," Web page, February 1999
[accessed 5 January 2000]. Available at http://www.ukoln.ac.uk/metadata/interoperability/.
This content downloaded from 185.2.32.121 on Sun, 15 Jun 2014 22:07:42 PMAll use subject to JSTOR Terms and Conditions
METADATA FOR MUSIC: ISSUES AND DIRECTIONS 209
the Dublin Core Metadata Element Set, as a means for the description and retrieval of music resources.8 Taking the lead in this area was the Performing Arts Data Service (PADS), a division of the Arts and Humanities Data Service in Great Britain, with many IAML members active in the project.9 The PADS
project raised several practical and theoretical issues of importance to the music community, some of which are being addressed by the developers of Dublin Core. Many of these and other issues are, in fact, common to a wide range of metadata schemes; but since the Dublin Core seems to be the
"metadata-of-the-moment," I will use it as the context for this discussion. Fast becoming the metadata scheme of choice for general library use, the
Dublin Core (DC) was developed within an international and interdisciplinary environment through the collaboration of various Internet communities, includ
ing librarians, computer engineers, text scholars, and most recently, the intellec tual property rights community. The Dublin Core semantics (i.e., content) was
kept simple because it was designed for authors of Web-documents to generate their own metadata.10 The development goal was to define a basic level of descrip tive data for Web sites that provided more information than just the keywords that were common in Web document headers, but was less complex than MARC
bibliographic records. This basic data could then be imported into any more
complex metadata structure (like MARC), and be used as the basis for building a richer description of the resource. The fifteen elements, many of which corre
spond, or map, to data in traditional catalogue records, are divided among three information categories: Content (Title, Subject, Description, Source, Language, Relation, Coverage), Intellectual Property (Creator, Publisher, Contributor,
Rights), and Instantiation (Date, Type, Format Identifier). There are currently no mandatory rules governing the Dublin Core Element
Set, which was designed for simplicity and flexibility. All elements are optional; all elements are repeatable; the order of elements is optional; and all elements can be qualified. Initially, the Dublin Core was designed to be syntax indepen dent, that is, no rules of grammar were mandated for the data so that the ele ments could be used in many different environments; but implementation of DC
projects required a syntax that could be used consistently on the World Wide Web without changes to existing Web standards and software. HTML <META>
tags were adopted by many early projects as the preferred syntax, and are cur
rently in the process of becoming a formalked standard; however, since
approval of the Resource Description Framework (RDF) as a structural stan
dard by the World Wide Web Consortium (W3C), XML has become the rec
ommended syntax for use with the more sophisticated RDF architecture.11
8. "Dublin Core Metadata Initiative," Web page, January 2000 [accessed 4 January 2000].
Available at http://purl.org/DC/index.htm. 9. "The Performing Arts Data Service," Weh page, 1998 [accessed 4 January 2000]. Available
at http://www.pads.ahds.ac.uk/padsWebNet98.html. 10. Stuart L. Weibel, "The Dublin Core: A Simple Content Description Model for Electronic
Resources," Bulletin of the American Society for Information Science 24, no. 1 (1997): 9-11. Also
available at http//www.asis.org/Bulletin/Oct-97/weibel.htm. 11. Eric Miller, "An Introduction to the Resource Description Framework," D-Lib Magazine 4,
no. 5 (May 1998). Available at http//www.dlib.org/dlib/may98/miller/05miller.html.
This content downloaded from 185.2.32.121 on Sun, 15 Jun 2014 22:07:42 PMAll use subject to JSTOR Terms and Conditions
210 FONTES ARTIS MUSICAE 46/3-4
Metadata Issues
While the fifteen basic elements of the Dublin Core have remained stable for three years now, virtually all applications of the Dublin Core elements have
adopted element qualifiers to accommodate more specific resource description than the basic model supports. This highlights the paradox that exists in the
DC community: on the one hand, the more elements and element qualifiers that are added, the less interoperable the data become; on the other hand, no one is really satisfied with the quality and level of description afforded by the basic element set. There are several important issues at the root of this prob lem, none of them new to the cataloguing community. These include:
• Lack of clear descriptions for multiple "states" of a resource, i.e., work,
expression, manifestation, and item. This issue involves the question: What state of the resource is being described?
Conflict between two different metadata functions, i.e., resource discov
ery and description. This issue involves the question: Can one core set of metadata serve two functions?
• Difficulty in describing aggregate resources. This issue involves the
question: At what level of granularity is the resource being described?
• Lack of mandated standards for data content. This issue involves the
question: Is authority control necessary for the description of Web-based resources?
As the PADS Workshop participants found, these issues present both theoret ical and practical problems that result in confusion when creating the metada ta record, and lack of precision in the description. Just when many specialized groups were beginning to despair of these issues being addressed adequately for their purposes, an unlikely ally appeared in the form of the Intellectual
Property Rights community.12 With this new and powerful group arguing for
many of the same descriptive details and standards as the library community, albeit for very different reasons, the Dublin Core developers began to sit up and take notice. The following discussion takes a closer look at the theoretical issues and some of the practical solutions that are being addressed by the Dublin Core metadata community.
State of the Resources Issues
Early on, those trying to apply the Dublin Core elements to the descriptions of actual resources had problems deciding on exactly what it was they were
describing. For example, the DC Date element seems straightforward enough, but exactly what date should be given? For a recording of the Mendelssohn
12. David Bearman, Eric Miller, Godfrey Rust, Jennifer Tränt, and Stuart Weibel, "A Common
Model to Support Interoperable Metadata," D-Lib Magazine 5, no. 1 (1999). Available at
http//www.dlib.org/dlib/january99/bearman/01bearman.html.
This content downloaded from 185.2.32.121 on Sun, 15 Jun 2014 22:07:42 PMAll use subject to JSTOR Terms and Conditions
METADATA FOR MUSIC: ISSUES AND DIRECTIONS 211
What is being Described?
Work
Expression
Manifestation
Abstract State
Tangible State Item
What is being Described?
Work
Expression
Manifestation
Abstract State
Tangible State Item
FIGURE 1
Octet, is this the date the Octet was composed? Or the date the performance was captured? Or the date the recording was originally issued? Or the date it
was digitized and mounted on the network? All of these dates represent valid
information that applies to different states of this resource's existence. The
IFLA (International Federation of Library Associations) model identified four
states of a resource as outlined in the Functional Requirements of the
Bibliographic Record,u These include the work, the expression, the manifesta
tion, and the item. Although there is still some confusion and perhaps dis
agreement over the model, it is generally agreed that the work and expression exist in an abstract state, while the manifestation and the item exist in a tangi ble state (see Figure 1). To give an example, if we choose a Mozart symphony as the work under consideration, its expression could occur either in sound or
in notation; a manifestation could be a performance or an edition, while the
item level could be represented by a specific CD or score.
In order for a resource description to be useful, it must be clear which state
of the resource is being described, and the descriptive data must be associated
with that appropriate state in some way. The current Anglo-American Cata
loguing Rules alleviate some of the ambiguity by requiring that the manifesta
tion be described in the body of the record, the work be described with access
points, and the specific item be described in notes.14 While reducing confusion
over what is being described, the end result is a single MARC record in which
some data elements describe the work, some data elements describe the par
ticular manifestation of an expression of that work, and some data elements
13. International Federation of Library Associations and Institutions. Study Group on the
Functional Requirements for Bibliographic Records. Functional Requirements for Bibliographic
Records. München: K. G. Saur, 1998. Also available at http//www.ifla.org/VII/sl3/ffbr/frbr.pdf.
14. Anglo-American Cataloguing Rules, Eds., Michael Gorman and Paul W. Winkler. 2nd edi
tion; 1998 revision. Chicago, IL: American Library Association, 1998.
This content downloaded from 185.2.32.121 on Sun, 15 Jun 2014 22:07:42 PMAll use subject to JSTOR Terms and Conditions
212 FONTES ARTIS MUSICAE 46/3-4
Resource Discovery vs. Description
Work High Recall Resource Discovery
Expression
Manifestation
Item
Resource Description
High Precision
Resource Discovery vs. Description
Work
Expression
Manifestation
Item
High Recall Resource Discovery
Resource Description
High Precision
FIGURE 2
describe a specific copy of the manifestation, all without clear indications as to which data are describing which state of the resource. This data model, which worked well for flat files with linear access such as card catalogues, has hin dered the development of new database structures for computerized cata
logues. Surprisingly, the basic Dublin Core data model, designed in a hyper text environment without the constraints of flat file access, fell victim to similar
inadequacies in describing the state of a resource, without the benefit of the guidance afforded by the cataloguing code. This problem could be solved by a one-to-one relationship between the descriptive metadata and the particular state of the resource being described. On the positive side, because of its sim
plicity, flexibility, and the small amount of existing Dublin Core legacy data
compared to the vast number of MARC records, the Dublin Core developers are able to work much more rapidly to solve the problem.
Resource Discovery and Description Issues
A tension has always existed between the two catalogue functions of resource
discovery and description. In order to discover or retrieve all relevant
resources, i.e., to obtain high recall, we need to describe at a high state, usu
ally at the level of the work or the expression. But in order to find the exact resource required, i.e., to obtain high precision, we need to describe at a much lower state, usually the manifestation or item level (see Figure 2).
As a core element set, the Dublin Core is really designed for high-level resource discovery, i.e., high recall. It is becoming increasingly clear, however, that once retrieved, searchers need metadata that can provide accurate and de tailed descriptions of resources in order to select the desired object efficiently. Thus, there is one faction of the Dublin Core community arguing for few and simple data elements to obtain the best recall and maintain a high level of inter
This content downloaded from 185.2.32.121 on Sun, 15 Jun 2014 22:07:42 PMAll use subject to JSTOR Terms and Conditions
METADATA FOR MUSIC: ISSUES AND DIRECTIONS 213
FIGURE 3
operability, while another part of the community is calling for more detailed
descriptions to obtain the best precision, even though this may mean sacrificing some interoperability. A solution to this problem may be found with new inte
grated library systems containing a variety of metadata databases that might use basic Dublin Core elements for the resource discovery aspect of a search, then link to records that provide richer resource descriptions when needed.
Level of Granularity Issues
Another persistent issue that presents problems for the creation and use of
metadata is the level of analysis for the description (often referred to as gran ularity) . Traditionally, bibliographic descriptions focus on the physical object and describe the whole entity at the summarization level. Many of these items are unified entities meant for linear access and there is no need to describe each section independently. The cataloguing rules do provide the option to
analyze contents and create separate descriptions when desired, and this
option is frequently used to describe the individual works on a sound record
ing. On the other hand, cataloguers rarely deal with description at the collec
tion level; that is usually the domain of archivists.
Electronic resources, however, present additional difficulties for deciding on the level of analysis for description. Web sites often consist of many indi
vidual parts that can stand alone, be used separately, and, therefore, be
described separately. Many sites consist of aggregates of images, texts, jour nals, or sound files (see Figure 3).
These sites might be described as aggregate collections, as single entities, as parts of single entities, or in any other way that may be useful to a searcher.
Determining the appropriate level of description and distinguishing between levels of description to insure intelligible assessment of search results can be
This content downloaded from 185.2.32.121 on Sun, 15 Jun 2014 22:07:42 PMAll use subject to JSTOR Terms and Conditions
214 FONTES ARTIS MUSICAE 46/3-4
problematic.15 Data models are needed that can represent both hierarchical and associative relationships with a fairly simple syntax and architectural structure. One solution to this problem is the fuller development of relation
ship metadata that can express clearly the associations and connections between separately described resources.
Authority Control Issues
Ultimately, the success or failure of the retrieval process for any metadata scheme rests on issues of authority control and data content standards, and these are certainly of great importance to music libraries. The ability of libraries to incorporate and use multiple metadata schemes in current library systems will depend on the compatibility of imported data with existing OPAC data. But even beyond the OPAC, authority control will play an important role in metadata interoperability.16 In order to perform a successful high-level search for music resources using Dublin Core elements, there must be some measure of authority control over titles and names. This raises the question: Are the overarching metadata principles of flexibility, interoperability, and
extensibility, which are so important to the metadata community at large, in conflict with the principles of authority control?
Flexibility has two aspects: flexibility in the amount of metadata and flexi
bility in the content. Flexibility in terms of the amount of metadata detail does not present problems for authority control. The varying levels of description prescribed by the AACR and the various Core Cataloguing Records already demonstrate this. Flexibility with regard to data content and form, however, does make authority control very difficult indeed. The descriptive rules, authority lists, and data registries used by different metadata schemes must be
identified, and options provided for metadata selection and mapping among dif ferent authorities.
In terms of interoperability, authority control becomes an issue when meta data from various sources and in various formats must be integrated and
manipulated for efficient use. While crosswalks allow data transfer from one metadata scheme to another, they focus on transfer architecture and syntax and do not address interoperability in terms of the standards used for data con tent and form. Thus, authority controlled data is not a requirement for the
physical transfer of the data, even though it would certainly improve the data
compatibility aspect of interoperability. Extensibility is not necessarily a problem for authority control, although, as
mentioned earlier, it does create other problems for interoperability. Some schemes are being augmented and refined to include more elements, more data qualifiers, and a more complex structure. In fact, many of the data quali fiers being added to metadata schemes such as the Dublin Core are designed
15. Willy Cromwell-Kessler, "Dublin Core Metadata in the RLG Information Landscape," D-Lib
Magazine 3, no. 12 (December 1997). Available at http//www.dlib.org/dlib/december97/ 12cromwell-kessler.html.
16. Sherry L. Vellucci, "Metadata and Authority Control," Library Resources and Technical
Services 44, no. 1 (2000), Forthcoming.
This content downloaded from 185.2.32.121 on Sun, 15 Jun 2014 22:07:42 PMAll use subject to JSTOR Terms and Conditions
METADATA FOR MUSIC: ISSUES AND DIRECTIONS 215
to support the use of authoritative data for names and subjects, and to indicate the authoritative sources for controlled information. Thus, extensibility can
actually enhance metadata authority control. Another factor for successful metadata authority control, and one that is an
integral part of the RDF structure, is reference to authoritative lists. In the RDF metadata model these authoritative lists, or data registries, provide meta metadata that may be used to 1) identify the metadata type from among vari ous schemes; 2) verify and validate a particular metadata's semantics and syn tax; 3) verify the agency that created the metadata; 4) verify specific standards used for data content and form; 5) identify specific resources; and 6) exchange data elements among metadata packages. These authority links are necessary not only because of the variety of metadata schemes, but also because of the
many different content standards and subject lists used in various disciplines. This data model, therefore, will require extensive building of authoritative data
registries and extensive relationship modeling.
Future Directions for Dublin Core and IAML
So, what are the developers of the Dublin Core doing to address these issues and problems? Four areas are currently under discussion for improvement, some of which will move the Dublin Core in the direction of a more stable metadata scheme.17 These include 1) formalization of the Dublin Core mainte nance process; 2) standardization of the element set; 3) development of a
mechanism for additions to data qualifiers; and 4) implementation of the Resource Description Framework.
Formalization and Standardization
Begun informally as a workshop on resource description, the Dublin Core
developers are moving towards providing an explicit process and organiza
tional structure for decision making. The new structure is headed by the
Dublin Core Directorate, which oversees the Dublin Core Policy Advisory Committee and the Technical Advisory Committee, and maintains the official
Dublin Core Web site. Working Groups may be formed to address particular
problems with elements (e.g., Date or Format), or clusters of problems (e.g.,
relationship modeling). The structure is similar to other agencies such as the World-Wide-Web Consortium (W3C) or the Internet Engineering Task Force
(IETF). Additionally, the Directorate created a formal process for ratifying
changes to the Dublin Core with input from various communities.
There will be some slight modifications to the Dublin Core prior to the next
stage of formalization, which is submission to both NISO (National Informa
tion Standards Organization) and CEN (Center for European Normalization) for status as an official standard. First, the data element definitions will be
reviewed to improve clarity and promote more consistent application. Second,
17. Stuart L. Weibel, "The State of the Dublin Core Metadata Initiative," D-Lib Magazine 5, no.
4 (1999). Available at http://www.dlib.org/dUb/april99/04weibel.html.
This content downloaded from 185.2.32.121 on Sun, 15 Jun 2014 22:07:42 PMAll use subject to JSTOR Terms and Conditions
216 FONTES ARTIS MUSICAE 46/3-4
the Dublin Core will be formatted according to ISO 11179, a standard for for
mally expressing the semantics of data elements in a standard manner.
Qualification Mechanisms
Data element qualification is the second area in need of improvement. From the outset, the many specialized communities such as performing arts and music made it clear that a mechanism to refine, or qualify, the metadata ele ments was essential for precision in resource description. The ability to add
qualifiers to the Dublin Core elements would improve the metadata scheme in a number of ways. It would 1) increase semantic specificity by allowing con trolled vocabularies or classification schemes to be identified and used for either description or as a formal browsing structure; 2) allow authority con trolled data to identify a name or title (e.g., DOI, SICI, BICI, ISBN, ISSN, ISWC, etc.) uniquely; and 3) allow definition of a formal substructure so that
compound values can be assigned to one element, e.g., Creator Value Name can have subelements such as e-mail address, affiliation, etc. Thus, data ele ment qualifiers will allow improved precision. It is the database structure and record architecture, however, that provide another level of meaning to these data by allowing specific data values to be identified clearly and associated with
specific functions, roles, or actions. This is the area that will benefit from imple mentation of the Resource Description Framework.
Resource Description Framework
The Resource Description Framework became an Internet standard in
February 1999. RDF is a set of conventions for expressing metadata that uses extensible Markup Language (XML) as an encoding standard and provides a framework for exchanging metadata. Resources are identified by a Uniform Resource Identifier (URI) and each resource has properties that describe it. By using the XML-namespace facility, RDF associates each property with its asso ciated metadata schema that defines that property, thus allowing data elements from different metadata schemes to reside together without confusion or con flict. The ability to specify metadata schémas in RDF makes it possible for
applications to access a particular scheme from a publicly accessible registry on the Web and retrieve the parsing structure and semantics of the element set. These link authorities have the potential to allow the development of rela tional systems that are only dreamed of currently with MARC. Figure 4 shows a very simple example of a metadata record that uses the RDF structure.
IAML and Future Music Metadata
If the Dublin Core is moving in the direction of formalization, standardization, increased descriptive data, and authority controlled data elements, in other words toward cataloguing data as we know it, are we just reinventing the wheel? I think the answer is "no," because although the semantics and syntax of Dublin Core are becoming more stable, they are designed for use within a structural data model that will allow the computer to do what it is good at: com
This content downloaded from 185.2.32.121 on Sun, 15 Jun 2014 22:07:42 PMAll use subject to JSTOR Terms and Conditions
METADATA FOR MUSIC: ISSUES AND DIRECTIONS 217
RDF Example 1
"IAML Presentation"
"Sherry L. Vellucci"
<?XML version»'1.0'?> <rdf:RDF xmlns:rdf="http://www.w3.org/TR/REC-rdf-syntax#"
xmlns:dc»"http://purl.org/dc/elements/1.0"> <rdf:Description rdf:about="URI:P">
<dc:title>IAML Presentation</dc:title> <dc:creator>Sherry L. Vellucci</dc:creator>
</rdf:Description> </rdf:RDF>
RDF Example 1
"IAML Presentation"
"Sherry L. Vellucci"
<?XML version='1.0'?> <rdf:RDF xmlns:rdf="http://www.w3.org/TR/REC-rdf-syntax#"
xmlns:do"http://purl.org/dc/elements/1.0"> <rdf:Description rdf:about="URI:P">
<dc:title>IAML Presentation</dc:title> <dc:creator>Sherry L. Vellucci</dc:creator>
</rdf:Description> </rdf:RDF>
FIGURE 4
putation and connectivity. This, in turn, gives us the potential to develop inte
grated library systems containing multiple databases that use multiple meta data schemes, employ sophisticated relational linkages, and provide unified interfaces without beginning from scratch.
So, where does this leave the music library community? Since anyone can
develop metadata and register it as an RDF schema, IAML's Cataloguing Commission could develop the metadata extensions and qualifiers necessary to describe music resources adequately, as the PADS project has already done to some extent, and declare these as an official RDF schema. LAML could
actively participate in this project by being the organization that officially main tains the schema. In addition, IAML could act as the clearing house for music metadata projects worldwide, allowing music librarians who are new to the metadata arena to benefit from the work and experience of their colleagues around the world.
This content downloaded from 185.2.32.121 on Sun, 15 Jun 2014 22:07:42 PMAll use subject to JSTOR Terms and Conditions