April 9-12, 2008
Montréal, Québec, Canada

The Meta Art Museum: Towards the Promise of an Open Collaboration Platform

Jenna Fleming, Museum of Fine Arts, Boston, Scott Shunk, Massachusetts Institute of Technology and Jeff Steward, Museum of Fine Arts, Boston, USA


With many collections available on-line, art museums face new opportunities in digital collaboration, service and scholarship. In this paper, we will provide an overview of one project to integrate images and data from the Museum of Fine Arts, Boston with Visualizing Cultures’ scholarly materials to create rich collaborative content on focused topics in Japanese history. While the technical approach is expected to evolve, it serves as a model for a relatively simple and “hands-off” method of collection integration that is easily maintained and easily extended, allowing the Museum to provide content for multiple on-line projects without significant extra investment. This paper will also discuss lessons learned and future opportunities for art museum collection integration, including the emerging standard of CDWA Lite.

Keywords: metadata, collections, collaboration, art museum, databases


Soon after the Museum of Fine Arts made its collection of more than 350,000 objects available on-line, MIT’s Visualizing Cultures team approached the Museum with a proposed project to integrate with the Museum’s on-line collection. Within a relatively short time, mechanisms had been built to allow a low-maintenance integration of data to be featured in Visualizing Cultures learning units on the Web.

Visualizing Cultures ( weds images and scholarly commentary in innovative ways to illuminate social and cultural history. Founded in 2002 by MIT Professors John Dower and Shigeru Miyagawa, Visualizing Cultures exploits the unique qualities of the Web as a publishing platform to enable scholars, teachers, and others to:

  • Examine large bodies of previously inaccessible images.
  • Compose original texts with unlimited numbers of full-color, high-resolution images.
  • Use new technology to explore unprecedented ways of analyzing and presenting images that open windows on modern history.

Visualizing Cultures has positioned itself as a nexus between the institutions that house image collections and the scholars who would like to use them for research purposes. Publishing on MIT’s revolutionary OpenCourseWare - making MIT courses freely available on the Web - Visualizing Cultures has worked with many institutions to negotiate on-line publication of images for educational purposes using a creative commons license. Visualizing Cultures offers a growing number of titles, referred to as “units.” This first pioneering set of units visualizes diverse aspects of Japan in the modern world. Units in development move into China and beyond.

The MFA is home to one of the finest and largest collections of Japanese art outside Japan, making it an important source of content and a key relationship for Visualizing Cultures. Now numbering over 100,000 objects, the Museum’s holdings encompass Buddhist art from as early as the eighth century; medieval narrative scrolls; ink paintings from the fifteenth century; masks for the stately No drama; remarkable swords and sword furnishings; unparalleled screens and scrolls, including important works from virtually every important school of painting from the sixteenth to the nineteenth century; and outstanding textiles. About 80 percent of the Japanese art in the MFA is made up of Japanese woodblock prints and books, including works acquired in Japan in the late nineteenth century and the fabled collection of John and William Spaulding, given in 1921. More recently, Jean S. and Frederic A. Sharf donated prints and photographs depicting the emergence of Japan as a modernizing nation, and Leonard A. Lauder presented the Museum with his collection of 20,000 postcards.

Figure 1

Fig 1: A selection of the MFA’s strong Japanese art collection, which includes over 100,000 objects.

Visualizing Cultures approached the MFA shortly after the development and publication  of its first unit, Black Ships and Samurai. For that unit, Visualizing Cultures cleared more than 200 graphics (all in the public domain) through more than 40 institutions, both public and private. The MFA offered the advantage of being a single institution with strong and relevant collections related to Visualizing Cultures’ scholarly interests. Working with one institution in this manner helped truncate the onerous process of IP rights and clearances.

The value of the collaboration between the two institutions was clear. Visualizing Cultures had a relationship with the Museum that allowed a very forward-looking collaboration to take place; that is, to augment the current understanding of the MFA’s vast collection through the work of noted scholars and to make that understanding available on-line through the ‘portal’ of the Visualizing Cultures framework. Visualizing Cultures is in the top 10% of all MIT OpenCourseWare sites (based on usage); with millions of unique visitors to OpenCourseWare each month, Visualizing Cultures’ reach and impact – and by extension, the Museum’s – is substantial. 

Data to Content: Transforming the Collection into On-line Education

The Museum provides both a searchable collection database and specific ‘collection tours’ which allow for a more selective and guided experience. Providing Visualizing Cultures with access to the collection additionally promotes deep and rich outlets for image-driven scholarship. The goal of Visualizing Cultures is to transform the manner in which scholars approach research when using images, and in turn transform the way  audiences view historical imagery. One of the key missions of Visualizing Cultures is to promote image-driven scholarship; in other words, the research and subsequent scholarly text is guided by the exploration of images and does not rely solely on textual information. There is an important distinction between an illustrated essay and a scholarly text based on images.

To do the research, analysis and interpretive work, the Visualizing Cultures team and lead scholar on the project sort through hundreds, if not thousands, of images, augmenting digital image searching with study in the Museum archives if needed. This process results in categories and groupings that begin to inform and shape the historical narrative for a unit. All supporting media at this point is digitized, allowing scholars to examine extreme details and delve even more deeply into the exploration of these often complex graphics.

Figure 2

Fig. 2: The unit “Asia Rising: Japanese Postcards of the Russo-Japanese War” uses images from the MFA’s Leonard A. Lauder collection of Japanese postcards. Images in the Image Database are returned in real time from the MFA.

The core content of a unit includes an essay and images, the visual narrative that expands themes from the essay, and other media components (video, animation or interactive content). As part of a unit, the databases that contain all the original images used to develop the unit content are created. This varies from 100 to 2000 graphics.

Once a unit’s image collection is defined, a very specific set of images is brought into the Visualizing Cultures Image Database (VCID). This is done live and in real-time; in other words, when a Visualizing Cultures user looks through a unit’s database, the resulting search is returned in real time from the MFA. The advantage is that this allows a very specific collection of images to be returned in a highly contextualized manner within a scholarly framework, an experience that clearly differs from a straightforward collection search. This provides users with a guided and deep way to explore a larger selection of materials on a subject such as the Sino-Japanese War.

The MFA is not the only institution to be collaborating with Visualizing Cultures in this manner. Visualizing Cultures also works with the Smithsonian Institution (the unit “Yokohama Boomtown, Foreigners in Treaty Port Japan” is one example of this collaboration). Other institutions that have engaged with Visualizing Cultures include The Hood Museum at Dartmouth, the Smith College Museum of Art, The Hiroshima Peace Memorial Museum, Allentown Art Museum, Bard Graduate Center, Chicago Historical Society, Chrysler Museum of Art, George Eastman House, Harvard University, Bishop Museum, Kobe City Museum, the Library of Congress, Nagasaki Municipal Museum, Nagasaki Prefectural Art Museum, New Bedford Whaling Museum, Shimura Toyoshiro collection, Shiryo Hensanjo-University of Tokyo, Smithsonian Institution Division of Photographic Resources, Tokyo National Museum, US Naval Academy Museum, US Naval Historical Center, White House Historical Association, Yokohama Archives of History, and Yokohama Museum of Art.

Opening Doors to Collaboration: The Open Knowledge Initiative (OKI)

The collaboration between the MFA and MIT’s Visualizing Cultures project is based on standards produced by the Open Knowledge Initiative (OKI), a consortium of MIT, Stanford and many other institutions. The Initiative develops and promotes specifications that describe how the components of a software environment communicate with each other and with other enterprise systems. OKI specifications enable sustainable interoperability and integration by defining standards for Service Oriented Architecture (SOA).

OKI identified an opportunity to develop a relationship with Visualizing Cultures in which OKI standards and functionality could be implemented and tested while allowing Visualizing Cultures to open access to vast archives of images from the MFA and other institutions.

The MFA was the first test case. OKI ‘hooks’ were built into the Museum database, allowing Visualizing Cultures to link directly to it with specific controls. This link opens  up the MFA content not only to the Visualizing Cultures database but also to any other environment that can use the OKI standards. Through this integration, users in other environments can search the MFA database (and any others integrated in this way) directly from the Web site they are browsing, instead of laboriously searching multiple archives one at a time.  This seemingly simple distinction begins to fulfill the promise of federated image searching.

How It Works: Technical Approach

The data integration between the MFA’s collection and Visualizing Cultures is simple; this has been one of its greatest benefits in terms of maintenance. A standard HTTP request is sent by Visualizing Cultures to the MFA Web server. The requested URL contains search parameters that execute a query against the collections on-line database. The results are wrapped in XML that conforms to the standard agreed upon by OKI for this project.

<?xml version="1.0" encoding="utf-8"?>










<title>A Young Man Dallying with a Courtesan, from an untitled series of twelve erotic prints Ukiyo-e print</title>


<date>about 1675-1685</date>

<medium>Woodblock print (sumizuri-e); ink on paper</medium>

<dimensions>Horizontal ôban; 26 x 36.8 cm (10 1/4 x 14 1/2 in.)</dimensions>

<creator>Hishikawa Moronobu










<title>Actor Tsutsui Kichijô in the Spear Dance Ukiyo-e print</title>


<date>1704 (Hôei 1), 11th month</date>

<medium>Woodblock print (tan-e); ink on paper, with hand-applied color</medium>

<dimensions>Vertical ôban; 53.5 x 31.5 cm (2in.)</dimensions>

<creator>Torii Kiyonobu I





[subsequent results omitted]



Fig 3: A fragment of unformatted XML data generated from the MFA database representing the kind of output used by the Visualizing Cultures site.

Setting up this system was as straightforward as using the existing queries and logic for the Collections Online portion of and redirecting the output to XML. The maintenance burden is light, requiring only a handful of fixes per year to conform with database changes. There is a small library of functions that retrieves data from the Collections on-line database. If there are structural changes to the database, or if OKI makes a change to the schema, the functions need to be updated to reflect those changes.

Overall, the system functions well, but there are some improvements that would be useful. Anomalies in MFA data accessed through Visualizing Cultures often result from holes in the MFA’s indexing of the data. Until the Museum can revisit and refine its data, it will be difficult to correct this.  An extremely useful enhancement would be a designation in between a scholarly description and a layman’s description of an object. For example, curators might not describe a painting as “Impressionist” because that is understood from looking at the image and considered to be unnecessary. However, since it isn’t possible to search data visually, general Web browsers will be unable to find Impressionist paintings unless they are already aware of applicable artists to search for. These descriptors could be addressed formally by curators or informally as keywords applied by other museum staff and the public (through folksonomy, in the latter case).

Future Directions and Related Integrations

Going forward, the MFA is interested in continuing to explore emerging standards and solutions in low-risk ways to help further the goals of projects such as Visualizing Cultures and the general goals of greater intra-museum collaboration and exchange. One of those methods is Categories for the Description of Works of Art (CDWA) Lite.

Thanks to pioneering work by the J. Paul Getty Trust, CDWA Lite is poised to become a popular XML schema for facilitating museum data exchange. The CDWA Lite schema has only nine required elements. This substantially reduces the burden on an institution when attempting to map data from its collections management system (CMS) to the schema. However, the reduced set of core elements is still substantive enough to enable meaningful discovery information.

The MFA is participating in two projects that make use of CDWA Lite:

  • A Research Libraries Group (RLG) project to build a tool to easily export data from The Museum System CMS to CDWA Lite.
  • A collaboration between the George Eastman House and the Metropolitan Museum of Art with the goal of aggregating data about daguerreotypes produced by the studio of Southworth and Hawes. The project team, including a representative from the MFA, is currently investigating the use of CDWA Lite for exchanging the core object information.

The MFA has great hopes for projects of this kind, but realizes that for many museums this goal remains unreachable. For some, digitization is still a future promise. For others, funding and resources to implement and maintain the tools that open the collection are often lacking.  Curators may be guarded about the data pertaining to their collection and worry about making seemingly incomplete data available to the public or to scholars. The MFA/Visualizing Cultures collaboration has helped to crystallize the shared team understanding that making available even incompletely described images can have great value to scholars and to the public, if only as a way to stimulate a larger community to participate in the task of description.

Lessons Learned

From this and related integrations (both current and planned), a number of lessons have emerged. Many relate to the positive benefits of this type of project, while others represent suggestions to avoid negative consequences.

Selecting an Approach – In assessing solutions to problems of data integration, a range of options exists, from simple to complex. We have seen great benefits from a very simple technical approach. A more complex one may or may not let us do more. There is much to suggest that a small but effective implementation may carry less risk than a large and sweeping one, especially in view of the pace of technology change. With the spirit of continued experimentation and evolving standards that marks the on-line collections space, a simple approach can offer a flexible solution that can be extended or overhauled without prohibitive cost. 

Data Normalization – A museum wishing to participate in this kind of integration must have an accurate electronic catalogue of its collection with clean, normalized data. For example, artists’ names must be standardized, and methods for communicating date ranges must be agreed upon across curatorial departments.

Long-Term Commitment - Once a museum opens the doors to its collections data in this manner, it must be committed to keeping those doors open indefinitely, or at least for the duration of the project at hand.

Willingness to Adapt – A participating museum should also accept that technological changes during the project’s timeframe may necessitate data migrations and upgrading of both hardware and software. If there is internal restructuring of how the data is housed, the organization will need to make sure the new structure will mesh with the access methods it has provided for the project.

Performance and Reliability - At the practical level, museums need to take into consideration the hardware they have and the hardware they’re capable of putting in place to reliably serve up this information. What loads are expected as a result of the new project? What are the potential bandwidth needs? Are those sustainable? Are qualified staff members available to run the servers needed? How reliable is the institution’s hardware, and to what degree has that been tested?

Large Files – In the current system, large files needed for some of the graphics in Visualizing Cultures are still “sneaker-netted” on DVDs between the two organizations. While it is not uncommon to find this happening, it is ripe for a standardized solution, whether that is a standards-oriented image extranet or another similar approach. A museum getting started with data integration may find that a solution to this part of the workflow is best as a second phase of work, somewhat separate from collections integration and more closely aligned with the needs of licensing departments. 

Can There Be a “Meta Museum”?

The adoption of a ubiquitous, open-source protocol such as OKI or CDWA Lite that provides a shared schema is a key step in more universal collection integration – both among museums and between museums and external projects. This promise of federated searching and data delivery has been a holy grail, to some degree, through the last several years of museum collection digitization.

As a collaborative team from a forward-thinking museum and a revolutionary technology institution, we are increasingly of the opinion that the path to this state of collection-sharing grace may be through evolution: by focusing on highly contextualized projects with clear ownership and outcomes, we build the smaller connections that, over time and with improving standards, become our shared virtual network of knowledge.

If a museum (with a digital project under its belt) can “open a door” to its digital collection using a standard data exchange method such as CDWA Lite or OKI, then tapping into one, some or many collections becomes far more technically feasible than ever before. The MFA or another museum could augment its on-line tools (such as custom collection building) by expanding the available artwork data to include other large collections. Any museum could become a meta art museum of sorts, in that sense, given the underlying organizational partnerships that would need to occur. An external project such as Visualizing Cultures could tap into a vertical slice of data, utilizing the sub-collections it needed from across organizations with accuracy and, ideally, with improved productivity and speed.  And while it may be controversial, an aggregator such as Google or iTunes could treat a museum database as it would any other rich source of content, returning results through a Web search engine or other delivery interface.

While we don’t pretend to know what the answer is (though we hope our experience may help provide data points), we believe there is an important benefit to engaging in this type of project-based integration for the purposes of learning about one’s own collection. With any project like this, a museum should create avenues for gathering feedback from the external institutions, with an eye to improving the data and responding to obstacles encountered by people accessing that data in new ways. Because MIT and Visualizing Cultures have a different perspective on the data than the MFA has, they may see flaws in data organization that the Museum is blind to or that its users have learned to work around. By gathering project feedback and establishing a plan to respond to it, a museum can make the data it exchanges with other institutions more accurate, complete and valuable for future collaborative projects.


CDWA Lite.

Museum Collections Sharing Working Group (host organization for the RLG Museum Data Exchange Project).

OKI and Museums.

Cite as:

Fleming, J. et al., The Meta Art Museum: Towards the Promise of an Open Collaboration Platform , in J. Trant and D. Bearman (eds.). Museums and the Web 2008: Proceedings, Toronto: Archives & Museum Informatics. Published March 31, 2008. Consulted fleming/fleming.html