Museums and the Web 2005
Screen Shot: What the user sees

Reports and analyses from around the world are presented at MW2005.

Expanding the Knowledge Base Managing Extended Knowledge at the National Museum of the American Indian

Michael Vulpe, i4i, Canada, and Jane Sledge, National Museum of the American Indian, USA


A Native American basketmaker was hired to teach a college class in Indian crafts. Each day she began and ended her class by having her students sing songs she had taught them, until some of the students began to complain that they would never learn basket making at that rate. So the teacher said, "Now we'll collect the materials for making baskets" – but it turned out that in order to do that she had to teach them still more songs, and the students became frustrated again. Finally, after they had collected all the materials and returned to the studio the teacher announced that she would now have to teach them still more songs in order to make the baskets. At this the students rebelled and demanded to know how to make baskets instead. To which the teacher replied, "Don't you know that the basket only represents the song, and that the songs are really the tradition that you need to make baskets? For a basket is only a visual representation of a song."

Web technologies have the potential to expand the depth and scope of the knowledge base by including other, often underrepresented, sources of knowledge. This paper explores a project underway at the National Museum of the American Indian (NMAI) that uses Web technologies to reach out to Native American communities and enables them to be active participants in the development of the knowledge base and in the sensitive care of Native American collections. The strategy goes beyond the simple 'feedback' forms on Web pages and relates directly to NMAI's mission to advance, in consultation, collaboration and cooperation with Natives, knowledge and understanding of Native cultures. The NMAI is testing a suite of Web infrastructure tools that offers Native communities opportunities to represent themselves in the knowledge base and locally manage knowledge about NMAI objects, media, and documents and related objects often stewarded by others.

Focusing on technical aspects, this paper will discuss the importance of a metadata server that provides a data front end to a virtual content aggregator to bring together previously unconnected collections databases. The metadata server, which can normalize an unlimited number of metadata namespaces, is front-ended by knowledge organization applications such as XML word processors and e-learning visual studios. It allows users to create and manipulate personalized metadata driven points of view (POV) on the aggregated collections. NMAI is testing the application of these tools in its collections research, public programs, and community services activities.

NMAI hopes that, by extending a level playing field using Web technologies, the knowledge base at the NMAI will serve as a public and educational resource that enables appreciation for the multiplicity of POVs of all subject matter experts.

Keywords: metadata, knowledge management, xml, collaborative content


The National Museum of the American Indian (NMAI) faces many challenges to effectively use its electronic information resources to meet its business objectives. The challenge discussed in this paper is how NMAI addresses the need to support collaborative interchanges of information by offering integrated modes of access to its own and other databases and incorporates, within its systems, appropriate additional knowledge gained through the collaborative processes.

Many projects successfully bring data together from diverse resources. (France chartered the development of national databases to hold cultural information in 1968 and Canada launched its National Inventory Programme in 1972). This challenge addresses the next step: what people want to do after they have access to the diverse resources.  The challenge includes ensuring that research, comments and expanded knowledge can be held in NMAI's systems. For example, a recent New York-based NMAI exhibit entitled, The Language of Native American Baskets: From the Weavers' View, featured more than 200 baskets from the NMAI collection.  None of the information prepared for the exhibit was entered into NMAI's Registration Information Tracking System because neither the system, designed to support the move of collections, nor the work-flow processes for museum record-keeping could support the information.  NMAI identified a potential solution in an emerging methodology: using a separate metadata tier as a mechanism of integration and a work and staging area for extended information.  NMAI asked Infrastructures For Information Inc. (i4i), an organization experienced in metadata management, to identify and define a metadata repository implementation to address the challenge, enhance the Museum's business processes, and contribute to its strategic goals.

NMAI contracted with i4i to produce a proof of concept project sufficient to demonstrate and evaluate, in controlled use, key concepts inherent in the use of an XML metadata repository to support on-going and collaborative content development, knowledge capture and knowledge representation through placeless document views.

This project is linked to other efforts: internal efforts such as NMAI's implementation of Collections Information System (CIS) and Digital Asset Management (DAM) software, and external efforts such as Indigenous Knowledge Management software (Hunter et al 2002), a project previously presented at Museums and the Web by Dr. Jane Hunter, Senior Computer Scientist at Distributed Systems Technology Center in Brisbane, Queensland. The software tools developed by Dr. Hunter et al enable authorized members of Indigenous communities to define and control a variety of sensitive care needs, accessibility, and reuse of their digital resources; uphold traditional laws pertaining to secret/sacred knowledge or objects; prevent the misuse of indigenous heritage in culturally inappropriate or insensitive ways; ensure proper attribution to the traditional owners; and enable Indigenous communities to describe their resources in their own words.

The goal of the project is to identify, define, and implement a proof of concept metadata repository implementation that can serve both as a "finding aid" for the museum to users of its various databases; for instance the CIS and DAM databases or public users of NMAI's Interactive Learning Center on the National Mall; and as a framework for collaborative development supporting NMAI's internal and external projects such as the Indigenous Knowledge Project.

The metadata repository is based on relevant industry standards, particularly on XML, for defining metadata and for managing content created in the collaborative environment. Commercial off-the-shelf (COTS) software is used, though some configuration and customization work is needed to adapt the software to the museum's particular content. The repository fits into the high-level system architectures set by the Smithsonian Institution's Office of the Chief Information Officer (OCIO) and the NMAI, and makes use of existing capabilities and services wherever possible.


Databases and other information resources proliferate within enterprises, and networking based on standard protocols such as TCP/IP and HTTP makes access to them broadly available, both within and among organizations. The challenge is to create a unified view or views of these resources to enable users to find what they need with a minimum of exploration, and to support software applications that can work with a set of databases without the need for separate interfaces to each one.  An emerging methodology to address this fundamental problem is the use of controlled vocabulary expressed in XML to create a normalized metadata tier to provide users with location, navigation, and content categorization management services across multiple independent databases. An example of this type of service is the WinFS component of the Microsoft Longhorn operating system, promised for 2006 but now delayed indefinitely. 

The Function Of Metadata

Simply put, metadata is data about data. Managed metadata can provide meaningful categories for content elements. These categories can then be used to create structured abstracts of the information held in the data stores below, and to present navigable views of the resources. There is a relatively long history of the use of a distinct metadata layer in document management and geographic information systems, but in these applications the metadata is drawn from a unified data store with a homogeneous data model and vocabulary. Advances in technology, specifically XML, which supports the design of very flexible data models and the interleaving of models, allow heterogeneous databases to effectively project themselves in metadata.. The projected metadata is a binary pair; an XML tag which defines 'aboutness’ and a content object which is the ‘subject’, are captured in a metadata tier that is data model agnostic. The net result is a consolidation of all of the participating databases presented in a normalized form and organized to support certain interactive services efficiently. (This approach is consistent with the evolving Web services model expressed in architectures like J2E and .NET. In these architectures, physical services and data stores have neutral (XML) abstract representations that are available over the Web for interrogation. The neutral abstraction masks complexities;,physical and logical, that need be resolved only once a decision to use a service or dataset is reached.)

A metadata repository integrates, to some degree, the contents of the participating databases into a common view for users (persons or software processes). The degree of sophistication of the consolidation is proportional to the sophistication of the normalized form.  

 MetaData and the Repository

To a database designer, metadata comprises the names and definitions of the components of a data model: tables, columns, relationships, constraints, etc.  We call this prescriptive metadata. The designers of object-oriented software hold a similar view. To the curator, cataloguer, or document manager, metadata comprise the values of pre-defined data elements that identify or classify an object, intellectual work or document; author, title, subject, provenance, etc. We call this descriptive metadata. A Google search on the term metadata will yield a result set in which both views are represented, but as two almost completely disjointed subsets.  Both views are relevant to a metadata repository. The following diagram illustrates their relationship:

Screen Shot: descriptive and prescriptive metadata

Fig 1: descriptive and prescriptive metadata

Metadata is built on data content. The data layer below the metadata repository represents a set of repositories holding digital content, in XML or other formats. Digital content may be in the form of individual files such as Word documents or organized in databases. Information processes and services, constrained by the security model, may also be included in this layer, because they too can be described in metadata. These resources are represented in the metadata repository by descriptive metadata, which must include an explicit identification (a data repository identity and some sort of address or link) in order to retrieve the content entity. Aside from this reference, the metadata repository is separate from, and essentially agnostic to, the data repositories; stronger couplings are possible, but increase the complexity of the metadata repository software for little gain in functionality.

The lower layer of the metadata repository is a type of directory, encoded in XML, containing data elements defined in the dictionary layer above, populated by values that either (a) describe individual information resources or (b) define classification schemes or categories used to organize views of the resource collection. In a sense, (a) represents cataloguing of various kinds and (b) represents organizational frameworks of various kinds – thesauri, topic maps, or simply arbitrary hierarchies of folders. The content units of the directory itself are XML instances.

The upper layer of the metadata repository is a type of dictionary that defines the data elements and their relationships, and is used to describe and relate information resources in the directory layer. In XML, the dictionary can be implemented as a schema or DTD or a set of schemas and/or DTDs that define the XML used in the directory layer and in any XML digital resources. A resource discovery tool would query the dictionary to be informed as to the meaning of the metadata used in the directory.

This two-layer model of a metadata repository provides several potential values not inherent in a single-layer model:

  • It can represent all resources. Without the dictionary layer, the directory is locked into a single representational model of the resources layer and is not easily adapted to new kinds of resources or new forms of organization.
  • It can dynamically support the creation of new digital resources; that is, the process of creating digital resources can be modeled in the dictionary layer and managed in the directory layer, so that the descriptions of resources and their organization are captured as they are created. This is particularly powerful when the new resources are assembled from existing digital resources and existing directory content.
  • It can easily support multiple views of the resources, organized in different ways for different uses or users. Directory data elements can be arbitrarily designated in the dictionary layer as carrying classification values so that associations between content entities and hierarchies of associations can be created as needed.

The content entities represented in the directory layer can be organized into various hierarchical views by associations built on common values in directory data elements. There may be other relationships between content entities, represented by cross-references or links embedded in the content itself. In principle, such links could also be managed in the metadata repository at the directory level; however, link management would only include XML content and would entail metadata server's being more tightly coupled to the XML content repository, significantly increasing its complexity. By maintaining a loose coupling between the metadata repository and the resource repositories whose content it represents, the metadata repository fits cleanly into most existing high-level architectures.

The metadata repository does not impose any restrictions on the data layer repositories; it only requires that the metadata about the content entities it describes be passed to it, or routed by it, as content entities are created or changed. An interesting aspect of this is that the metadata repository does not need to represent all of the content in any or all of the data layer repositories. This allows metadata management services to be developed over selected content, enabling an evolutionary approach to implementation.

A Starting Point For NMAI

The starting point for the NMAI was an XML repository that created an aggregated metadata view of data from Smithsonian Institution museums and offices. The data includes information from the Smithsonian's American Art Museum Web site, Native American Portraits from the National Portrait Gallery's Catalogue of American Portraits, a combination of the Smithsonian Institution Library and Northwestern University's Edward Curtis Collection images, the National Anthropological Archives glass plate negative portraits of Native Americans, and NMAI's collections. The idea was to bring together an interesting subset of information that could provide users with insight about Native American materials at the Smithsonian and illustrate how a metadata repository addressed NMAI's objectives – managing knowledge capture and facilitating collaboration.

The metadata view is delivered in way that allows the museum's external constituencies to collaborate with the museum to build new information products and expands the factual and interpretive knowledge base about the collections.  The metadata repository is not the interface but the enabler of diverse interfaces. This addresses key business needs articulated in Director West's Strategic Plan 2002-2004:

Develop and produce multi-dimensional exhibitions that reach out to visitors to connect them with our message, utilize new and emerging technology tools, link them to Native sources of knowledge, and provide intergenerational learning experiences.

(NMAI Strategic Plan)

Knowledge Capture & Collaboration

The problem with knowledge capture is that new information is infrequently returned, managed, or preserved in a meaningful way in museums' collections systems. Knowledge about the collection is developed during the process of research projects, publication planning, and exhibit design. This knowledge is captured and held in documents. Unfortunately much of this 'added' knowledge does not return to the collection information system; it remains in the curatorial files and publication documents. It may be that this information should NOT be returned to the collections system as it is tangential to the collections data, or of a different genre, or owned and held by other communities or institutions.  Staff should be able to choose to include information (records, documents, or media) in the collections system, the digital asset management system, the metadata repository, or through the metadata directory link to related information in other museum, library, and archive systems. We do not have to stuff all our expanded knowledge into a single system.

Screen Shot: The NMAI structure

Fig. 2: The NMAI structure tells the user that NMAI thinks first of object type, then where it came from (geography), then the tribe and the artist/maker.

The problem of collaboration has both technical and human aspects. While the human aspect of getting people to work together is separate issue, the technology aspect that can help make it as easy as possible for people to access subject materials is central to the objectives of the metadata repository project. An aspect of the project is defining and managing an electronic 'public space' wherein collaborative activity takes place. That public space brings together information resources that are a necessary part of an informed discourse as well as contributions to the discourse. Documents are central to the discourse. They are a primary knowledge capture and collaboration medium. Authors put their thoughts and knowledge on to 'paper' – documents. These are distributed to the collaborative community for review, comment, and edit. It can be said that in an asynchronous digital world, the document is the medium.

XML is a core technology for enabling access to document content. An XML repository for the storage, and ultimately, management of publication materials, supported by XML content creation and manipulation systems, addresses the problem of knowledge capture and collaboration. Specifically, XML is used to identify knowledge developed during document development and editing: the 'aboutness’ and ‘subject’ pairing of metadata in the metadata repository. The XML knowledge document is abstracted as metadata that populates the metadata repository and when projected to the normalized form is used to define and organize a public space for the collaborative activity. (A further application of this software tool will take place within a project called Indigenous Knowledge Initiative).

A Pilot

Native communities seek local control over cultural heritage and language preservation efforts and desire to speak for and represent themselves to the museum's public. To meet these objectives, NMAI and other partners are considering a strategy, still in the formative stages, called the Indigenous Knowledge Initiative which brings together community, education, and technical experts, computer research projects, and electronic software resources outlined below. These resources will improve the ability of individual Native communities to strengthen culture and language at home while, at the same time increasing their ability to learn about other cultures, educational practices, and languages. The Initiative is centered in concepts of Native community control and Native Ways of Being approaches to teaching and learning, developed by Native people from around the world

The Initiative will create a distributed network built around culture and language preservation and continuance in Indigenous communities.  The Indigenous Knowledge Management System (IKMS) is a software tool developed by Smithsonian Fellow and Senior Computer Scientist Dr. Jane Hunter, at the Distributed Systems Technology Center (DSTC) in Queensland, Australia. Jane Hunter started with the premise of enabling Native communities to use multimedia technologies to "record and preserve significant aspects of their cultures, including languages, ceremonies, dances, songs, stories, symbols, design, artwork, tools, costumes, historical photographs, film, videos, and audio tapes." (Hunter et al 2002). In concert with NMAI, Jane Hunter sought to develop electronic mechanisms to enable traditional owners

to define and control the rights and access to their resources, in order to: uphold traditional laws; prevent misuse of Indigenous heritage in culturally inappropriate or insensitive ways; and receive proper compensation for their cultural and intellectual property. … [NMAI staff believes] it is essential that Native communities describe and contextualize their culturally and historically significant collections, in their own words and from their own perspectives (Hunter et al 2002.)

KMS software is designed for use by any community willing to work with NMAI and the DSTC to develop the expertise necessary to implement the software in their community. Member colleges and universities from the American Indian Higher Education Consortium (AIIHEC) and World Indigenous Network Higher Education Consortium (WINHEC) reviewed the Indigenous Knowledge Management software, and two colleges are involved in the project as beta testing sites. Little Priest Tribal College in Nebraska is developing a three dimensional tour of the Ho Chunk Society Historical Museum located on the campus.Working with experts from NMAI, middle school students from the community's schools are taking photographs of the local museum's artifacts and displays. They are also meeting with elders and historical society members as part of the effort to develop annotations for each object included in their tour. Local objects will be supplemented in the virtual museum by working with NMAI to photograph Ho Chunk artifacts stored at NMAI's Cultural Resource Center located in Suitland, Maryland.

NMAI's challenge is to manage and make accessible appropriate information resulting from this Initiative. NMAI will use its metadata repository to provide a staging ground and gathering site for materials generated in the Indigenous Knowledge Management software. The metadata repository 'integrates' commercial collections management and digital asset management software products as well as XML-MARC. This unified view enables tribal members to explore music or oral recordings, for example, in museum collections or on-line tribal archives from around the world, along with records about the instruments and records about objects being created while the music is played or sung.

Tom Davis, a contributor to the First Voices Project, writes,

Robert Peacock, former Chairman of the Fond du Lac Band of Ojibwa Indians in Minnesota and now the President of White Earth Tribal and Community College, used to dream that he could go over to the tribe's small museum, and then use computers to explore Anishinaabe artifacts and items and information available from Canada, Europe, or wherever they might be found.  (First Voices)

A preliminary view is that the Indigenous Knowledge Initiative has 7 major activity stages, suggested by Jane Hunter in the Software Tools for Indigenous Knowledge Management resource (Hunter et al 2002).

Screen Shot: Seven activity stages

Fig. 3: Seven activity stages of Indigenous Knowledge Initiative

They are as follows:

  • Generation of a digital information request
  • Submission and review of a digital information request
  • Organization of resources and the provision of these resources to the native community
  • Attachment of rights
  • Application of annotations
  • Disposition of information by the native community
  • Review of information by the NMAI
  • Disposition of information by the NMAI

The following diagram shows the basic workflow of the spiral, with the metadata repository providing the initial application support for NMAI processes and Jane Hunter's Indigenous Knowledge Management System providing the application support for community use.

Screen Shot: Basic workflow

Fig. 4: Basic workflow

At NMAI the information / documents are available in a shared Web space which has the 'look and feel' of Windows folders with documents. The default folder structure is the NMAI organization model. That default structure is populated with publications, collections, and media asset documents. The metadata from the collections documents, the media asset store and the publications documents is used to create the model and the documents 'appear' in folders because of their metadata. Change the metadata in a document, and it automatically appears in the appropriate folder. This allows content to appear in multiple places simultaneously. The physical constraints of organization are removed; objects position themselves based on their metadata properties.

individual folder structure

 Fig. 5: This individual folder structure tells the NMAI that the user thinks first of the tribe, then the season, then the theme, then the material

Folders, being in effect XML documents with metadata, act as documents do. Because documents and folders position themselves logically rather than physically, the user is able to create multiple axes of understanding without having to worry about the physical placement of objects in the structures – they appear automatically.

The shared Web space appears to the user as an extension of the Window's file system. Documents are accessed on users' computers from the shared Web space. Saved documents pass through the metadata server, their properties extracted and the metadata views updated appropriately. They are then stored in an XML document repository or returned, in their native form, to the collections information system or media asset server as appropriate.

These logical folder structures, which are in fact XML documents viewed through the prism of Window's Explorer, are available to Web browsers. This paradigm may be easier for casual users and viewers to accept and to navigate.

Screen Shot: What the users see

Fig. 6: What the users see

Note: The creation of XML representations of collections records and media assets is an extract of the current database(s) into XML form. The XML is used to populate the metadata repository and is the neutral intermediate form that the new CIS system will support. It is also the basis for data interchange between the metadata asset system and the CIS.

Expected Outcomes

New knowledge sources, such as researchers or tribal visitors, can add knowledge to the collections information documents. That knowledge is expressed in:

  • direct supplements to the collections documents
  • references made from the publications documents to the collections documents
  • alternative organization models for the information
  • Publication products (Web sites, exhibits, books, Web casts, video-conferences and educational resources) incorporate and attribute the newly acquired knowledge.


NMAI is developing a technical infrastructure to support knowledge gathering and collaboration to manage, maintain, and present information for use in a variety of museum activities. No one system solves all the challenges. NMAI chose to acquire a variety of systems to meet the needs of specific functions and to integrate these systems at a higher level. Just as no one system solves all the challenges, no single museum, library, archives, or information store can hold all the information and knowledge.  Museums are interdependent with libraries, archives, universities and larger constituent communities. Unlike many libraries and archives, museum activities expand the existing knowledge base.  Museum systems are not static; museum systems are dynamic and ever expanding. The systems must enable communities to converse, collaborate, generate, share, and comment upon information held in multiple systems. We must be able to link resources across multiple systems, add information and comment, cooperatively engage with external constituents and communities, and develop new interfaces and products.


First Voices Project., accessed October 26, 2004.

Hunter, Jane, Bevan Koopman, and Jane Sledge. September 2002. "Software Tools for Indigenous Knowledge Management." Accessed October 26, 2004.

NMAI Strategic Plan. National Museum of the American Indian, Strategic Plan 2002-2004.

Cite as:

Vulpe, M. and J. Sledge, Expanding the Knowledge Base Managing Extended Knowledge at the National Museum of the American Indian, in J. Trant and D. Bearman (eds.). Museums and the Web 2005: Proceedings, Toronto: Archives & Museum Informatics, published March 31, 2005 at