VLMA: A tool for creating, annotating and sharing virtual museum collections

Introduction

§ 1    Artefacts and artworks have been used traditionally in the teaching of archaeology, biology, meteororology, palaeontology, and zoology (to name a few), as well as history and the fine and applied arts. Some of these academic disciplines—namely archaeology and history, especially the history of art—have found a place under the umbrella of Classics, due to (a) the increasing sophistication with which material culture—real ancient objects, including documents, tools, and even 'art'—is employed by ancient historians and (b) its longstanding although less sophisticated use as 'background scenery' for the study of literary works. The growth of research in Classical material culture is reflected in its popularity in the classroom. At least in England, the number of material culture modules in Classics departments are on the rise.^[1] If not for the increase in local travel costs and insurance and simultaneous decrease in time available, one might expect a concomitant rise in academic visits to archives, museums and collections. Given the real life conditions mentioned above—constraints on time, travel, and expenditure—visits of students from their middle teens are steadily decreasing. Yet even if a teacher is lucky enough to cobble together a 'field trip' he must rise to the challenge of himself preparing students for a meaningful visit and preferably follow up the visit by incorporating the museum/archive materials into the fabric of the module studied. Since 1995 web materials have provided us with a plethora of materials for the virtual study of museum collections but hardly any tools to enable us to make efficient use of them. The Virtual Lightbox for Museums and Archives (hereafter VLMA) uses P2P technology to enable students, researchers, and other teachers to create their own collections of museum and archive materials from a range of museums and archives, then to annotate, organise, and present their selections to their students or fellow researchers. All this without having to invest in software, CDs, DVDs, or site licenses. VLMA likewise gives data providers—museums and archives—a cheap and nonexclusive means of publishing their visually rich electronic databases on the web, in a manner that will make them easily navigable and useful to their visitors.

§ 2    Data providers (curators, archivists, and administrators) and users (students, researchers, and other teachers) have equal shares in a 'datasharing problem': how to analyse, synthesise, utilise, and package visually rich collections in a meaningful and efficient manner. VLMA addresses this problem. This paper focuses on the teacher's role in this cycle. The teacher—even more so than the researcher—is in the critical position: if the teacher cannot deliver the material then the material remains unused. After a discussion of the pedagogic use of visually rich museum and archive collections, we will explain VLMA's functionality, structure, and means of implementation, and report on its status quo. The same problems bedevil all involved with visually rich datasets, but in this paper we focus on our primary expertise with museum datasets.

Sharing museum data at the turn of the millennium

Figure 1: Entrance to the Ure Museum (October 2005).

§ 3    Frustrations in curating and teaching with a small museum collection, namely the Ure Museum of Greek Archaeology (at the University of Reading), at the turn of the millennium, inspired the creation of VLMA: a simple, browsable, searchable federation of museum and archive datasets. The Ure Museum is and has been, since 1922, a department collection that gives '… life and variety to the study of Greek History,' according to its founders, Annie and Percy Ure (See Figure 1).^[2] It is housed within the Department of Classics at the University of Reading, which owns and manages it. Funding from the Arts and Humanities Council (AHRC) has just enabled the first overall redesign of the museum since its installation on Reading's Whiteknights campus ca. 1960. The new display contextualises life in ancient Greece and Egypt, presenting artefacts in thematic units, such as 'Household', 'Education', and 'Body beautiful'. A 'timeline' on the north wall of the Museum provides an overview of Egyptian and Greek pottery forms as they developed from prehistory to the Late Roman period. Together with a map, drawings, text, and the vases themselves, the timeline provides the background a visitor needs to understand how most of the objects in the Museum came about. Such was Annie Ure's aim in amassing a collection of ancient Greek pottery—to show her students all representative fabrics, styles, and techniques of ancient Greek pottery. As with any museum display, however, the vases are embedded here in a context that the curators and designers have imposed upon them. If one wants to know more about one of those fabrics, styles, or techniques, one must look beyond the Ure Museum. Annie Ure, as indeed most educators and curators in the last millennium, would regularly present students and other visitors with comparable material (a.k.a. comparanda) through visual aids: illustrated catalogues and other books; stacks of postcards; box after box of photographic negatives; and, at best, expensive museum quality photographs (almost always black-and-white, given the relative clarity of b/w as a photographic medium). Most of these were and continue to be precious enough to guard under lock-and-key but, except for the slides, hard to share with groups of students. It is a testament to Annie Ure's energy and resourcefulness that we have so many of these in the Ure Museum, but they were and are expensive and increasingly hard to acquire (because of copyright issues). The internet might have improved the accessability of comparanda, but copyright restrictions are more of a threat than ever (cf. a recent review from the British Academy: Copyright and Research in the Humanities and Social Sciences ) and the cost of high quality images continues to rise. Because of copyright restrictions we can only publish digital versions of our own collections, not the supplementary materials amassed from other collections.

§ 4    In the ongoing internet revolution many large museums have digitised their collections although most smaller museums have yet to rise above the costs of knowledge, staff time, server storage space, and conservative archivists who see digitisation as a trendy nuisance that will endanger the collections in the short run and distract from the problem of long term storage. Few large museums are fully digitised. The British Museum's Compass portal, for example, provides text and images of only a small fraction of the Museum's holdings, while staff train teams of volunteers to enter information about the rest of the collection into its digital database. The images provided on Compass represent a very small percentage of the photographic holdings for the featured objects, let alone the hundreds of thousands objects in storage. The British Museum is a few steps ahead of the crowd, however: Compass provides high quality information in a user friendly format and has done so since June 2000. Most digitised museum databases are on intranet servers, as at the Louvre Museum in Paris, for the use of museum staff and the occasional academic visitor, while web pages enlighten visitors as to 'select works'. The widespread use of intranets is a default response to the challenges of user interfaces, integration, and a widespread anxiety about giving the data away 'for free'.

§ 5    Before a museum publishes its data to the world it usually spends an inordinate amount of staff time (and that of more expensive consultants) creating a user friendly interface with an interpretative framework. When one considers that the average Museum exhibits less than 25% of its holdings (the Ure Museum displays closer to 50% of its holdings), the scale of such work is a massive hurdle. There are also the exclusion problems inherent in imposing an interpretative mask and choosing 'select works'. Who is to know what each visitor is looking for in a museum? One solution to the latter problem is being addressed by a consortium of American museums in the Art Museum Social Tagging Project (http://www.steve.museum/).

§ 6    Both within single museums and among related museums there is the problem of integration. Many of the larger museums have multiple datasets that are difficult to integrate. While London's Science Museum and others are making headway on integrating their own complex datasets (see http://www.sciencemuseum.org.uk/collections/about/collections_info_sys.asp), CIDOC, the documentation arm of the International Council of Museums (http://www.cidoc.icom.org), proceeds with Martin Doerr's CRM or Conceptual Reference Model (http://cidoc.ics.forth.gr/). The CRM provides a common and extensible semantic framework onto which one can theoretically map any cultural heritage information. The mapping is the obstacle here: who does the mapping, when, and where? The skills required for such mapping are still beyond the average museum IT staff, let alone curators.

Teaching with a museum collection: finding a context

§ 7    Following its achievement of registered museum status in 2001, under the Re:source scheme (see http://www.mla.gov.uk/documents/musreg_eng.pdf), the Ure Museum has made an effort to open up its resources and facilities to the University as well as the wider community, and aimed to digitise the entire collection. The result of largely volunteer effort is the Ure Museum Database or Ure DB (http://lkws1.rdg.ac.uk/cgi-bin/ure/uredb.cgi).^[3]As the Ure DB neared completion in 2003-2004 we began to research and plan the implementation of user interfaces. In this effort we reconsidered why, besides publicity, museums go online. We encountered the usual problem of second-guessing the users: Who were they? How would they want to view and use the materials? What would they use them for? We had initially thought that a variety of user interfaces designed for specific age groups would optimise usage but we were wrong. Feedback obtained through a Widening Participation project in 2003, sponsored by South East Museum, Library and Archive Council (http://www.semlac.org.uk/casestudies_urefull.html) encouraged us that users at all ages and academic levels longed for (a) the greatest flexibility, to go beyond the ‘tombstone’ information provided about objects in traditional museum labels, and dig as deep as they wished into the database; (b) images that provided multiple views as well as details of artifacts. Users craved such access both to be able to find and study, on the one hand, one object in detail, and to have sufficient information to be able to compare, on the other hand, two or more objects. The database, which presents each object through text and image, simply opens up the museum contents, not merely the displayed objects, to all potential visitors. Virtual access to the Museum should help students – whether at Key Stage 2 (aged 8-12), AS- or A-level (aged 16-18), University (aged 18+), or beyond, to prepare for and follow up on their visits to the Museum or to prepare projects based on the collection (or parts of it).

§ 8    The Ure Museum database gets its users, and especially teachers who use it, only so far. For pedagogic purposes, and indeed any enhanced understanding of the artefact, some contextual information is necessary and the relevant context is usually found beyond the collection. Two examples of this phenomenon that are particularly relevant to the Ure Museum are the need to (a) (re)combine distributed objects or (b) combine distributed assemblages of objects. Ancient art, like all archaeological materials, are fragments of the cultures that they represent, and usually fragments of the objects from which they come (Smith 2000). Often a scholar might find a fragment of a sculpture or vase in one museum that joins to a similar piece in another museum. Dyfri Williams has done just that with an Archaic Greek vase fragment in the Ure Museum (inv. 26.2.1: http://lkws1.rdg.ac.uk/cgi-bin/ure/uredb.cgi?rec=26.2.1) that joins a dinos (bowl) attributed to the painter, Sophilos, which is housed in the British Museum (inv B601.26 or B100: Williams 1983, 34). So access to the fragment on the Ure DB gives visitors only a glimpse of the whole, and to see the more significant parts of the vase, one has to have access to the corresponding piece in the British Museum (it is not yet included on Compass).

§ 9    The same Archaic fragment is also part of several distributed assemblages of objects. For example, someone interested in the works of Sophilos would wish to consult all of the 91 works attributed to or signed by that artist: the single fragment on the Ure DB, and the more impressive objects in the British Museum, Greece's National Museums, and elsewhere. These are fortunately brought together, albeit in limited form, on the Beazley Archive website (http://www.beazley.ox.ac.uk/), which for reasons of ‘copyright’ only reaches a restricted audience that has applied for and been granted password access. Alternatively one might also be interested in studying the other vases and vase fragments that, like our Sophilos dinos, were found at the great Archaic Greek trading post, at Naucratis, in Egypt. British Museum staff are now researching and (ultimately) gathering the Naucratis assemblage. Meanwhile the relevant material is mostly unavailable on line. This is but one example of how a small collection such as the Ure Museum is dependent, for both teaching and research, on comparisons of its holdings to those in other museums and archives.

§ 10    The inherent interest in comparing museum artifacts is perhaps the best reason for all museums, especially those with limited collections, to publish their collections on the web. Whereas big museums such as the British Museum can comfortably expose its own collections on the web without running out of 'context' (to which they own copyright) the smaller museums, researchers, and especially teachers are at the mercy of others for context. The normal situation for teachers offering thematic presentations is that of Annie Ure, discussed above: they collect objects (images and other information) from a variety of other collections. For presentation purposes, the average teacher ‘cuts and pastes’ artefacts (images or data) and their contexts, either electronically, into power point or a webpage,^[4] or physically, in creating an old fashioned 'handout'. The resulting powerpoint, webpage, or handout might then successfully present the results of their research and creative ('cut-and-paste') work, but is usually a dead end, by which we mean a resource that cannot easily be adapted by the teacher, students or others. Teachers need a way of building virtual collections that address their own pedagogic needs and interests and those of their students but also give the students some material on which to expand. This is where e-learning and e-research should intersect, but have not yet met. It is the researcher who demands more context and provides it where it hasn't been previously given. The teacher and student can and should play this role as researcher—and would eagerly do so—if they have access to contextual materials, without the massive expenditure and time associated with travel. That is, the web could and should provide the consumer (teacher or researcher) with artefacts and contextual materials through tools that would allow them to collect, reuse, and even enrich the resources themselves. We have called this process of reuse and enrichment ‘syndication metabolism’ (see Figure 2).

Figure 2: Diagram illustrating syndication metabolism, or the process of collection, reuse, and even enrichment of online resources, showing relationships between data providers and consumers.

Visual resource comparisons

§ 11    Just as it is impossible to answer research needs within one museum, it is impossible to answer research needs within one museum website. Differences in presentation from website to website severely limit this potential for researchers, let alone lay audiences, to cross the institutional boundaries. There is also the familiar difficulty in maintaining references to off-site data. The semantic web is envisioned as a solution to heterogeneous web presentations, but the visual element is often left out of these discussions (Berners-Lee 2006; Berners-Lee 2001; Fensel 2002). Useful comparison of museum materials relies on the opportunity to compare images as well as the other information (usually textual) about the artifacts that they represent. Image comparison technology is widely available to professionals who prioritize images in their work—architects and designers, other artists, and technicians, to name a few. The Virtual Lightbox (http://mith2.umd.edu/products/lightbox/applet.html), an open source project developed by the Maryland Institute for Technology in the Humanities (MITH: http://mith2.umd.edu/), for example, provides users with an opportunity to collect various images and then to sort and / or modify them in relation to each other, as on a traditional (non-digital) lightbox or light table. VLMA has adapted this useful open source applet to the museum and archive sector by enabling the user to retain data and metadata (text concerning the image as well as the object represented by the image) while viewing and sorting the images. The retention of metadata is crucial to the applicability of lightbox technology to museums, for the purposes of teaching and learning, despite the fact that museum websites are as guilty as any of presenting images with little or no textual reference to the artifacts that they represent, or in a manner in which viewers could inadvertently lose (close) the textual reference.

The VLMA Solution

Federation and data provision

§ 12    VLMA enables users access to objects—whether museum artefacts or archival documents—via their images and text descriptions, as well as relevant metadata (e.g. when and where a photo was shot) (see Figure 3). Yet it considers usability to be as important as access, so that sticky metadata, annotation, and user selection are inherent in the system. We discuss here VLMA's provision of data and services through federation—an important factor that permits functionality going far beyond the bounds of the tool itself.

Figure 3: Diagram explaining the relations of item objects to (textual) data sets, images, and relevant metadata.

§ 13    As a means of providing data, VLMA enables the federation of disparate collections: any collection that wishes to publish their data in VLMA may do so, simply by running VLMA on their server. Like any P2P client VLMA can list the other collections that are simultaneously published in the VLMA interface and bring them to the attention of users (see Figure 4).

Figure 4: VLMA displaying available 'friends' (‘serverlist’).

Federation means that a user is able not only to navigate through the collection hosting the VLMA, but also through 'friend’ collections, which in turn offer access to other friends, and so on. This provides three key advantages. Firstly it makes the user aware of relevant collections previously unknown to her. Secondly it simplifies navigation: the user interface is homogenous across collections, regardless of the heterogeneity of the data. Finally, it allows for searching across collections. Although search criteria remain limited, as is the case in the current phase of development (with relatively little overlap between collections and lack of homogeneity between data fields), it is possible to use VLMA to search for key terms over several different collections simultaneously and to return the results in a single set.

Syndication: Empowering the consumer

§ 14    VLMA presents an alternative solution to existing web portals, whether single institution systems like the British Museum's Compass, or federated solutions like European Cultural Heritage Online (ECHO: http://echo.mpiwg-berlin.mpg.de). It also works with and encourages metadata initiatives, such as OAI (http://www.openarchives.org/) and CIDOC (http:// www.cidoc.icom.org). We have called VLMA a 'portlet' solution (Smith 2005), meaning a resource whose function is to provide access to distributed information systems at the level of the consumer of information rather than (as with other solutions), at the level of the provider. Because VLMA works as an applet or as a Java Web Start application, it takes the portlet metaphor a step further. It frees the portlet from the server environment, turning it into a free-floating user-agent. The VLMA is thus an independent consumer-peer that interacts with a network of consumers and providers. Like a portal, it is capable of accessing a diverse set of collections. Unlike a traditional portlet it is not a component of any one portal, but rather a collection point for data chosen by the information consumer, not the information provider (See Figure 5).

Figure 5: Diagram showing services shared between consumer and provider.

This approach differs from most attempts to integrate content in that it leaves to the consumer the tasks of collecting, organising, and reusing data. For this to work the data provider must simply 'syndicate' the item objects in the collection, that is, describe where to find them and what tools to use to access them. VLMA uses RDF (Resource Description Framework: http://www.w3c.org/RDF/), which consists entirely of aggregation points for material—visual and metadata resources, as well as tools such as RDF parsers and datastores—already available on the web (Lasilla and Swick 1999). We are in the process of creating a tool that will allow data providers to generate and publish the required RDF by inputting a simple tab-delimited file that identifies where image and metadata resources are.

§ 15    The syndication process described above avoids the biases inherent in HTML presentation in two ways. First, all collections look the same to the consumer: they are simply bags of item objects that can be collected for local reuse. Second, by syndicating services rather than contents of collections, we avoid a presentation based on ontology. The VLMA is blind with respect to the material being accessed and therefore never prefers or prioritises one over the other.

VLMA services

Figure 6: Diagram explaining relationship of consumer and provider regarding VLMA services.

§ 16    What the user sees in the VLMA window is a group of collections, each of which exposes its collection as a group of access services (see Figure 6). The three basic services are Browse, Search, and Friends. The Browse service allows browsing of the entire contents of a collection. The Search service enables structured searching through a collection. The Item Object is, however, the central unit of transaction. Each Item Object consists of the cluster of images, data, and metadata fragments attaching to an object (see Figure 3). This can then be 'collected' by the user into her own collection for further use.

§ 17    One of the collections visible to the user is his own ('MyCollection'), which offers the same services as the collections of data providers, with one exception. In the place of 'friends', there is a lightbox service, which contains objects that have been collected to the lightbox panel and may be viewed and compared there. The federating function of the Friends service is taken over by the Search service, which differs from the Search service in non-local collections in allowing the user to do a federated search over all of the collections that the user has discovered on the net by using the Friends service. An annotation service enables the user to add her own comments locally, which is a crucial step in reuse, either for purposes of teaching or research. With or without annotation, however, the user may export her own collection, as discussed below.

The lightbox panel and export services

§ 18    Tools for examination and comparison of retrieved data, whether textual or visual, are also provided in VLMA. A lightbox panel enables the user to arrange, resize, and manipulate her own selection of images (see Figure 7). As the lightbox naturally does not edit the images itself, they can also be restored to their original forms at any time. Text comparison tools are more basic but allow for the hiding of empty attributes as the metadata is stored internally in XML.

Figure 7: VLMA window, showing VLMA services, including lightbox panel.

§ 19    Exported data may be stored in several ways: as state, in machine-readable form, and in presentation format. The lightbox does the first two by exporting to XML, the W3C-recommended format for data storage and transfer (http://www.w3.org/XML/). The file can be saved to the user's system, enabling her to recreate her own personal collection in future, regardless of which host site is being visited. The format is formally closed and text-based, making it easily readable by other programs, as well as by human interpreters. We have enabled export to Impress, the presentation component of Open Office. (Those currently wishing to export to MS PowerPoint can do so simply by exporting from Open Office.)

Progress to Date

§ 25    A successful first development cycle in 2004-5 saw the development of the portlet application, along with a simple implementation of a data-hosting server and the syndication of the complete Ure digital collection. The second cycle of development has just reached completion and was intended to make various minor changes to the functionality of the portlet, to introduce VLMA as a Java Web Start application, as well as syndicate collections at our home institutions (Reading’s Typography collections and Herbarium, http://www.herbarium.rdg.ac.uk/, as well as the UCLA/Max Planck Institute’s Cuneiform Digital Library Initiative or CDLI at http://www.cdli.ucla.edu/index_html). This process brought up many of the ‘administrative’ problems which such a project is likely to engender, ranging from the simple unavailability of data that had been promised by third parties, to an overly large database that cannot be exported from its proprietary system without paying consultants large amounts. Whilst it is always tempting to put such issues down as ‘teething’, or Murphy’s Law, they have in fact been useful indicators of the typical state of such data. As a result, one of the primary goals of the third development cycle will focus on developing software to make the process of syndication as simple as possible.

§ 26    User testing was undertaken during both the first and second phases of development. Volunteer students who had not previously used VLMA were asked to perform a variety of predefined tasks using the software on several different browsers (IE6, Firefox and Safari). No direct support was given other than the online manual. They were then asked to rate how easy each task was using the tool and provide any further comments. Feedback was generally very positive, although the artificial nature of the tasks made judgment difficult, i.e. participants often had a tendency to ask whether they were doing the 'right' or 'wrong' thing, as opposed to remarking on how intuitive the interface was. While it was felt that the interface could be improved in a various ways (especially the context menus), all participants felt that the VLMA provided a useful way of interrogating and engaging with the database.

Conclusion

§ 27    VLMA presents users and providers alike with benefits of cost reduction, ease of manipulation, and transparency, all of which are crucial for public organisations such as schools, universities, museums, and archives. VLMA reduces costs for data providers, as well as users, in offering a completely free extension to any database, on any system, that makes the database instantly available to online users without risking the danger of being held hostage to proprietary software. Beyond having no installation costs, it adds value to the collection itself, providing services that enable users to explore not only items in the collection, but also the relationships those items have with others in collections the world over. As artefacts are rarely of great value in themselves but rather in what they can tell us of patterns in culture (material and otherwise) it is clear that the value of a multitude of collections is greater than the sum of its parts. By increasing the ease with which users can access the data they seek, VLMA encourages them to make more use of it, in terms of both the quantity of items browsed and the depth of information provided. This in turn will motivate collections and their funders to invest in documenting and publicising more of their collections. Transparency is increased as all information relating to a specific object is linked to it directly and managed only by the collection owner. As a result, the danger of altered items masquerading as the original or metadata being arbitrarily withheld (other than by the provider) is lessened.

§ 28    This brings us to perhaps the most intellectually encouraging advantage of VLMA, its 'hands-off' approach to metadata. While we consider metadata models to be highly beneficial, it is increasingly difficult for providers to reach consensus on international standards for museum data and its presentation. VLMA works in parallel to such standards initiatives, providing a method by which collections implementing multiple standards can be browsed simultaneously, without exclusion.

§ 29    VLMA challenges the status quo, in encouraging museum and archive data providers to extract their data from proprietary, complicated, or otherwise closed systems and to deliver it freely with tools that are easy to learn. We have begun to syndicate collections at our home institutions on VLMA in order to pilot VLMA’s capacity for federation, but have yet to prove its full potential as a teaching and research tool, because of the lack of homogeneity across our collections. In its third phase of development we will pilot VLMA’s viability as a research tool across related collections, with the cooperation of sympathetic data providers.