VADS information standards review

Tony Gill, Catherine Grout and Louise Smith
20 March 1997

	Introduction
	Background
	About the Authors
	A domain-specific review of relevant standards for networked information discovery

Introduction

This review of domain-specific information standards has been produced for participants of the 1997 Visual Arts, Museums and Cultural Heritage Metadata Workshop, organised by the Visual Arts Data Service in partnership with the Art, Design, Architecture and Media Information Gateway, the Scottish Cultural Resources Access Network and the Museum Documentation Association. It is part of a series of workshops being organised under the auspices of the Arts and Humanities Data Service and UK Office of Library and Information Networking, and is funded by the Joint Information Systems Committee of the UK's Higher Education Funding Councils.

Background

Information standards development is not an organic process which occurs in a vacuum, or an empirical exercise free from the influence of political and cultural factors—it is usually tied to the perceived needs of the day, and occasionally may even look to the achievements of yesterday. In the cultural field, we are often at pains to stress that our information standards are not technologically-driven, but this is only possible up to a point—frequently it is the development of new technology that creates new opportunities for applying our priceless information, and these in turn drive the need for new standards initiatives.

In order to make these initiatives more manageable, it’s been common practice to focus on a particular view of the available information, for example categorised either by function or by subject. The functional approach is illustrated in the Getty Information Institute’s Object ID Project: Core data attributes that could be agreed by cultural bodies, law enforcement agencies and insurance companies in their endeavours to retrieve stolen artefacts. Initiatives taking the subject-based approach include the Van Eyck Project, the focus of which is flat art, and the five information service providers under the umbrella of the Arts and Humanities Data Service (AHDS). This approach is not without its pitfalls, however, when one considers that the definition of an object (as being archaeological as opposed to fine art, for example) often has nothing to do with the inherent nature of the object itself, but is rather more related to a human viewpoint. It is possible, and indeed often desirable, to have more than one ‘subject view’ of the same object.

The diversity of disciplinary approaches is clearly illustrated in the field of terminology standards, where the emphasis in recent years has shifted to the use of thesauri and other linguistic resources to enhance information retrieval, irrespective of original cataloguing practices. Whilst there is still considerable work to be done in terms of developing terminology resources at the specialist level (at the time of writing the MDA supports terminology working groups focusing on archaeology, ethnography, medicine and transport), there is an equal effort required to ensure that such resources map into an overarching conceptual framework.

SPECTRUM: the UK Museum Documentation Standard has been implemented in a number of museums on a variety of platforms (including manual paper-based systems), thereby fulfilling the documentation requirements of the Museums and Galleries Commission’s Registration Scheme. The strength of SPECTRUM lies in its reinterpretation of documentation as relating to all the functions of the museum, not solely that of cataloguing. However, SPECTRUM is necessarily non-prescriptive in terms of the structure of the information that is captured and stored, which although key to its acceptance and adoption by the community, has inevitable implications for interoperability across diverse systems.

It is also becoming more widely acknowledged that there is a need to handle diverse (and often conflicting) interpretations. After all, in the field of cultural information the ‘right answer’ is often either elusive or manifold. In response, standards professionals are looking to the complex information capture afforded by newer methodologies such as object-oriented modelling and semantic network processing. It is no coincidence that the Documentation Committee for the International Council of Museums (CIDOC) is in the process of translating (a misnomer; rather it is a reworking) its relational data model to an object-oriented one.

The requirement to access distributed data has resulted in a number of projects but few standards; access to distributed databases is none too difficult if they are homogeneous. The current challenge is accessing heterogeneous information resources. The Consortium for the Computer Interchange of Museum Information (CIMI) has been working towards that goal for a number of years, and has implemented the CHIO pilot project utilising SGML and Z39.50, technical standards that are also being employed by the flagship EC project Aquarelle. Clearly, what is of interest to the cultural heritage sector is the potential of these standards if adopted more widely; recent work in Aquarelle has focused on separate DTDs for the so-called ‘built’ and ‘movable’ heritage, and the latter was heavily influenced by the CHIO DTD.

Metadata is a relatively new (and poorly-defined!) term for an old concept—the descriptive information about an object that assists in it’s discovery. It just so happens that the object itself is also information, hence the meta prefix.

The broad aims of the 1997 Edinburgh Workshop are to ascertain the extent to which generic (i.e. not discipline-specific) metadata standards can be applied to the information resources of the visual arts, museums and cultural heritage communities.

Like all standards initiatives, the workshop will be judged a success if it manages to build on the consensus of what’s gone before and feed into the consensus of what comes after. And indeed it is this very process which gives us pointers to the future, where standards feed standards (e.g. MDA Data Standard Þ SPECTRUM), standards feed projects (e.g. CIMI DTD Þ Aquarelle), and standards are embedded in off-the-shelf applications (e.g. SPECTRUM and AAT Þ Multi MIMSY 2000).

But these standards are not an end in themselves; they aim to enable information sharing across boundaries, whether the boundaries are subject-based, geographically-based or linguistically-based. The need for cross-boundary standards, built from a foundation of consensus from the ground up, has never been greater.

About the Authors

The Standards

It is worth remembering that few, if any, of the following standards are discrete; many have been based on what’s gone before and/or have been mapped one to the other. It is in this tradition that the VADS metadata workshop will be following; focusing on the specific requirements for metadata standards for the community whilst at the same time ensuring that existing standards feed into any resulting work.

A domain-specific review of relevant standards for networked information discovery

	Dublin Core and Warwick Framework
	IAFA Templates
	Categories for the Description of Works of Art
	Van Eyck
	VRA Core Record
	CIDOC Standards
	SPECTRUM
	MDA Data Standard
	CIMI DTD
	SCRAN Data Standard
	ObjectID
	MARC
	ICONCLASS
	Terminology Resources

Dublin Core and Warwick Framework

The Dublin Metadata Core Element Set, or Dublin Core, is a set of 15 descriptive elements used to provide a simple means of describing networked electronic information resources to aid more effective discovery and retrieval.

The Warwick Framework is a conceptual container architecture that can contain multiple metadata packets.

At the time of writing (March 1997), the draft definition of the 15 element set is as follows:

The 15 elements and their meanings have been developed and refined by a group of librarians, information and subject specialists, through an ongoing consensus-building process that has included 4 international workshops to date and the active Meta2 mailing list.

The first workshop, organised by OCLC and NCSA in March 1995 at OCLC’s headquarters in Dublin, Ohio, saw participants arriving at a consensus on the need for a simple Internet resource description record to aid information discovery. It was agreed that the record needed multidisciplinary support, and should also be very simple so that useful descriptions could be created by information providers without the need for training. The result was agreement on a set of (then) 13 elements for describing ‘Document-Like Objects’, all of them ‘repeatable, optional and extensible’.

The second workshop, held at the University of Warwick in April 1996, was organised by OCLC and UKOLN. This workshop aimed to broaden the consensus and promote interoperability across languages and subject disciplines, by acknowledging that the Dublin Core would not meet the diverse resource description needs of all communities, nor should it attempt to. The result was the Warwick Framework, a conceptual container architecture for holding multiple diverse ‘packets’ of metadata. Proposals for implementing this framework have been based both on an SGML DTD and the MIME architecture.

The third workshop in the series was held in Dublin, Ohio in September 1996. Organised by CNI and OCLC, it aimed to establish a core element set specifically for describing digital images, to ascertain whether the Dublin Core could be extended for this purpose. After much debate, it was agreed that most images were, in fact, Document-Like Objects, and therefore the Dublin Core Metadata Element Set could, with a few modifications, be usefully applied. After an intense period of debate on the Meta2 mailing list, a consensus emerged, along with some refinements to the element definitions and two additional elements (bringing the total to 15).

The fourth Dublin Core workshop, organised by OCLC, DSTC and the National Library of Australia, was held in Canberra, Australia in March 1997, where participants succeeded in arriving at a broad consensus on the need for an optional method for refining the meaning of the 15 Dublin Core elements through the application of three additional qualifiers—dubbed the Canberra Qualifiers—TYPE, SCHEME and LANG. These can be used to refine (as distinct from extend) the semantics of the 15 individual elements. Progress was also made in the application of the COVERAGE element.

Although simplicity is one of the fundamental tenets of the Dublin Core metadata initiative, there is a significant body of opinion within the stakeholder community that the ability to refine the semantics of the element set, through the addition of optional qualifiers, is essential to allow effective deployment of the Dublin Core for resource discovery.

For example, the value 759.2 (British Painting) in the SUBJECT element is effectively meaningless unless it is qualified as referring to the Dewey Decimal Classification system. Similarly, the elements DATE and COVERAGE have little meaning or value without qualification of some form; for example, does the date 07-03-97 refer to the seventh of March or the third of July, and which century is it in?

To address these shortcomings, the qualifiers TYPE, SCHEME and LANG have been agreed upon:

	TYPE allows the refinement and clarification of element content such as, for example, dividing DATE into Creation Date, Date Last Modified, etc. It can also be used to identify logically grouped items of information.
	SCHEME refers to an existing classification system, coding scheme, glossary or thesaurus, and is used to identify the meaning of an element's content.
	LANG refers to the language in which the contents of the metadata is written.

A separate working group has been set up to investigate registry issues for the standardisation of qualifier values, such as the logging of ‘acceptable’ coding schemes and their recognised abbreviations.

The formal acknowledgement of the requirement for qualifiers in Canberra paves the way for practical deployment of the Dublin Core.

The ability to embed structured descriptive information in WWW pages has always been recognised as a key strategic application for the Dublin Core: The following sections outline the syntax for embedding Dublin Core metadata without qualifiers in HTML, some of the reasons behind the development of the Canberra Qualifiers, and two recommended syntaxes for embedding Dublin Core metadata with Canberra Qualifiers in HTML.

The simple syntax in which Dublin Core (without the optional Canberra Qualifiers) is embedded in HTML relies upon the HTML <META> tag and takes the form:

<META NAME = DC.ElementName CONTENT = "value string">

For example:

<META NAME = DC.creator CONTENT = "William Shakespeare">

The <LINK> tag can be used to refer back to a registry of element definitions held at OCLC:

<LINK REL="SCHEMA.dc" HREF="http://purl.org/metadata/dublin_core_elements#subject">

The extended syntax described below is the first of two recommendations from the Dublin Core group, and is considered by the group to be sub-optimal; However, the preferred syntax requires changes to the HTML DTD that have not been agreed by the W3C at the time of writing. The syntax presented here conforms strictly to the HTML DTD from version 2.0 onward.

The qualifier is appended to the element name using a full stop (period or dot) as a separator:

<META NAME = DC.ElementName.TypeQualifier CONTENT = "(LANG = Language) (SCHEME = SchemeName) element value">

For example, using just the SCHEME qualifier:

<META NAME = DC.subject CONTENT = "(SCHEME=DDC) 759.2">

Using both TYPE, SCHEME and LANG:

<META NAME = DC.date.created CONTENT = "(SCHEME=ISO 639) (LANG = en) 1997-03-07">

The Dublin Core homepage is at http://dublincore.org/

IAFA Templates

Also known as ROADS Templates or WHOIS++ Templates

The IAFA templates were originally devised by the Internet Anonymous FTP Archives working party of the Internet Engineering Task Force, as a simple means to catalogue the contents of FTP archives. The records produced using the templates are stored as simple attribute-value pairs (i.e. the name of the attribute or field, followed by a separator and the field value), usually as flat ASCII text files.

There are a number of different types of IAFA Templates, used for describing different types of resources; most of the templates are used for describing individual files (documents, images, software package etc.), and use the same basic set of attribute (or field) values. There are also a number of templates for describing other types of resources, such as services, logical archives and mirrors.

Some of the attributes (fields) within each template can be repeated; these are known as variants. It is also possible to create clusters of attributes to describe USERs (i.e. people) and ORGANIZATIONs. These clusters can then be combined with a role prefix and embedded within other template types.

The attributes (fields) used in templates that describe individual files (e.g. DOCUMENT, IMAGE, VIDEO, FAQ etc.) are shown below:

<1> The intention of this field is to define the category of the object. For example, in the case of documents it could be "Technical Report", or "Conference Paper" and the name and date of the conference at which the paper was presented. It may also be something like "General Guide" or "User manual".

<2> Objects are often available in several formats. For example, documents may be in PostScript, ASCII text, DVI etc. For images this may be GIF, JPEG, TIFF etc. Format should be specified in MIME type syntax and semantics where possible (See RFC1521).

The following table shows the attributes in the USER cluster:

The following table shows the attributes in the ORGANISATION cluster:

Early implementations of the IAFA Templates for their intended purpose, the description of FTP archives, include Nexor’s ALIWEB search service and the Linux Software Map.

They are also employed by a number of other initiatives, most notably the ROADS software (Resource Organisation And Discovery System) used by a number of the Access to Network Resources projects of the UK’s Electronic Libraries Programme, for example ADAM (Art, Design, Architecture and Media Information Gateway, http://adam.ac.uk/), OMNI (Organising Medical Networked Information, http://omni.ac.uk/) and SOSIG (Social Science Information Gateway, http://sosig.ac.uk).

Categories for the Description of Works of Art

This section is based on the AITF web pages available at http://www.getty.edu/research/institute/standards/cdwa/index.html

Produced by the Art Information Task Force (AITF), the Categories for the Description of Works of Art are guidelines for formulating the content of art databases. They articulate an intellectual structure for descriptions of objects and images: in this sense they constitute a schematic representation of the requirements and assumptions implicit in the practice of the discipline of art history.

By providing a single, encompassing framework for descriptive information about works of art, the Categories are intended to enhance compatibility between diverse systems that wish to share art information. Such a standard will contribute to the integrity and longevity of information transmitted across networks and moved to new systems. Above all, it will give researchers consistent, reliable access to information stored in a variety of systems in far-flung places.

The framework defined by the Categories represents information that is suitable for sharing and exchange. The context for such an exchange might include preparing a catalogue for an exhibition, doing research for a scholarly essay, or locating works of art to see on a study trip. While much of the information defined by the Categories may also be found in local museum collections management systems, the task force decided that the intricacies of collections management such as object tracking are best defined by local museum professionals.

The AITF Categories for the Description of Works of Art focus upon "movable" objects and their images, including paintings, works on paper, sculpture, ceramics, metalwork, furniture, design, performance art, and so on, from all periods and all geographic regions. Although the Categories do not deal with the built environment, the depth and richness of the information the Categories articulate may serve as a model for descriptive standards in a range of fields in the humanities, including architecture.

Through its survey and review process, the task force identified 26 distinct kinds of information found in descriptions of works of art. While art historians may not automatically think of content in terms of categories of information, categorisation is in fact implicit in much of their work. The traditional catalogue entry reflects discrete, recognisable groups of information usually distinguished typographically for easy access by the reader. In a typical exhibition catalogue the artist, title, date, owner, provenance, measurements, materials and techniques, and current location are clearly separate groups of information. The AITF simply extends that process to the full range of information found in descriptions of works of art.

To make this categorisation neutral and applicable to many areas of study, the AITF has deliberately chosen broad category names. Thus, for example, the Creation category provides for both known and unknown individual artists, as practised in the study of European art, and for tribal and cultural group identification as used in African or Oceanic studies. The Creation category also accommodates all the other circumstances surrounding the creation of a work of art such as date, commission, and place.

Categories indicated as contain the minimal information that is needed to uniquely identify a work of art; these constitute "primary access points" in an automated environment. Of course this is not meant to imply that only the minimal amount of information recommended should be recorded about any given work of art.

Each category, in turn, may contain discrete items of information that a scholar expects to find about works of art. The sub-categories within each category identify and define those specific bits of information, grouping them together in a logical way. The sub-categories essentially describe the specific items of information that would be included in either a narrative or fielded description dealing with the category. As on the category level, sub-categories indicated as constitute primary access points for the minimal information needed to uniquely identify a work of art.

Van Eyck (Visual Arts Network for the Exchange of Cultural Knowledge)

*the inclusion of a visual matching software within the Van Eyck system which allows the user to search on the visual characteristics of an image as well as its text record

It was decided that a minimum record for art image data should be divided into three entities. The Artist or originator of the original work which is defined as the Object and the Image or images derived from it—which exist as a reproduction in a given archive. This lead to the definition of the following relational structure consisting of three tables with one to many relationships.

N.B. For copies of detailed Project reports on the Core Record please apply to Catherine Gordon, Witt Library, Courtauld Institute, Strand, London.

Visual Resources Association (VRA) Core Record

A full description can be found at http://www.gsd.harvard.edu/~staffaw3/vra/vracore3.htm

CIDOC Standards

	The CIDOC Archaeological Sites Working Group published a Draft International Core Data Standard for Archaeological Sites and Monuments in June 1995.
	The CIDOC Ethnography Working Group published the International Core Data Standard for Ethnology/Ethnography in September 1996. This includes the results of a survey of ethnographic documentation practice and a core data standard for ethnographic collections.
	The International Guidelines for Museum Object Information were published by CIDOC in June 1995.
	The CIDOC Multimedia Working Group published Introduction to Multimedia in Museums, November 1996.
	The CIDOC Data Model is available as an Entity-Relationship model. The intention is to produce future CIDOC data models using O-O methods, techniques, and formats. While this effort is under way the current CIDOC Relational Data Model will be available, although it will not be modified or enhanced. The scope of the current relational model concentrates on the documentation of museum objects and their provenance; the O-O model will enable more specific collections management and research information to be included.

In addition the AFRICOM Inventory Working Group has completed work on a handbook of standards, the aim of which is to establish common documentation standards for use by museums throughout Africa to facilitate the development and exchange of information.

A full description can be found at http://www.cidoc.icom.org/guide/guide.htm

SPECTRUM: the UK Museum Documentation Standard

In the context of the Edinburgh Workshop it may be useful to look at the overview of the Units of Information provided by SPECTRUM, since this groups attributes together according to the type of information they provide:

MDA Data Standard

Revised in 1991 the standard was originally developed in the 1960’s, when it focused entirely on issues of cataloguing. It was extended in the 1980’s to include some aspects of collections management, and in its latest edition consists of approximately 160 fields, which are categorised as being either Primary, Group or Common

CIMI DTD

Thanks to Richard Light for this information

CIMI (the Consortium for the Computer Interchange of Museum Information) is a group of museums and museum bodies which have come together to further the cause of pooling and exchanging museum information. CIMI aims to make progress at both a theoretical and a practical level.

CIMI’s current major project is Project CHIO (Cultural Heritage Information Online). This project aims to deliver at least 10,000 records of objects and information about folk art as a searchable online resource. This will include the full text of exhibition catalogues, wall texts, etc. as well as images and the more traditional museum database records.

CHIO aims to set up a demonstrator which will support different ‘points of view’ relating to this information. The ‘general public’ PoV will reflect the knowledge and interest of a typical museum visitor. The ‘museum’ PoV will reflect the knowledge and interest of a museum professional.

CIMI decided at an early stage in Project CHIO to use the SGML standard (Standard Generalised Mark-up Language; International Standard ISO 8879:1986) to mark up the textual resources involved in the project (exhibition catalogues and wall texts). In order to do this, it had to adopt or develop a Document Type Definition (DTD), a set of rules that defines the allowed structure of SGML documents. SGML requires that every valid document should have an associated DTD, and since no DTD’s had been developed for museum information of any sort, it was decided that a new DTD was required.

What follows is a list of the ‘atomic’ access point concepts defined for the project’s SGML DTD:

SCRAN Data Standard

The SCRAN (Scottish Cultural Resources Access Network) Data Standard has the following fields which have been mapped to Dublin Core and SPECTRUM:

Object ID

Machine Readable Cataloguing (MARC)

MARC (ISO 2709/ANSI Z39.2) was first deployed in 1968, and is undoubtedly the oldest metadata format in existence. Originally developed as a standard data record structure detailing every aspect in the storage of bibliographic data (down to the specification of the physical tolerances of the magnetic tape it is stored on!), it has been modified over the years to provide a rich and flexible structure for the storage of a wide range of information types, including the description of digital resources (the 856 fields).

The key benefits of MARC are derived from the fact that it is the standard bibliographic format for libraries around the world. However, relatively uncoordinated development of the standard over the years has resulted in a situation where there are now many different ‘flavours’ of MARC in use around the globe. Work is currently underway to harmonise these different versions as the UNIMARC standard, which is expected to be completed in 1998.

MARC has been used for storing internet metadata information in the INTERCAT project http://www.oclc.org/news/oclc/cataloguing/type.htm, and is the basis of the Z39.50 Profile for a Digital Library http://www.loc.gov/z3950/agency/ An SGML DTD implementation has been produced, and can be found at http://www.loc.gov/marc/marcdtd/marcdtdback.html.

ICONCLASS

Terminology Resources

For a comprehensive index of resources see WordHoard on the MDA web pages at http://www.mda.org.uk/wrdhrd1.htm