copeland.indd The Academic Library as a Gateway to the Internet 251 Context and Meaning: The Challenges of Metadata for a Digital Image Library within the University John Attig, Ann Copeland, and Michael Pelikan To be successful in the university environment, a digital library must be able to integrate content from faculty and students, as well as traditional library sources. It must have a robust metadata structure that can accommodate and preserve a variety of discipline-specific metadata while supporting consistent access across collections. As part of the Mellon-funded project, the Visual Image User Study at Penn State, a prototype centralized digital image delivery service was created and explored. In creating a metadata schema for the project, the authors anticipated both a wide variety of content and users across many disciplines.This schema employed three very differ- ent standards (VRA Core Categories, Dublin Core, IMS Learning Objects Meta-data).The project validated the need for highly individualized content, the importance of individual faculty collections, the need for editorial interven- tion to supplement and modify contributed metadata, and the importance of addressing discipline-specific vocabularies and taxonomies. igital libraries are developed within a context, and that con- text determines the character of the design. Most digital libraries are organized and administered by libraries, but many exist within a uni- versity system. In this context, the teach- ing and research functions of the univer- sity are determining factors. Members of the university community are rapidly becoming accustomed to using digital ob- jects in their teaching and research. They have their own content (both digital and not-yet-digital), and they seek additional content relevant to their work. Faculty members are also likely to be users of university course management systems that provide an electronic environment for the classroom experience and which make digital learning objects available to their students. If a digital library is to be successful in this environment, it must integrate these sources of content and these systems into its design. This paper seeks to validate one aspect of this argument—the design of an ap- propriate metadata infrastructure—based on the authors’ participation in a project to study the needs of users of images in an academic environment. As part of this John A ig is Monograph Cataloging Librarian/Authority Control Librarian and Associate Librarian; e-mail: jxa16@psulias.psu.edu; Ann Copeland is Special Collections Cataloging Librarian and Assistant Librar- ian; e-mail: auc1@psulias.psu.edu; and Michael Pelikan is Technology Initiatives Librarian and Assistant Librarian; e-mail: mpp10@psulias.psu.edu; all at Pennsylvania State University Libraries. 251 mailto:mpp10@psulias.psu.edu mailto:auc1@psulias.psu.edu mailto:jxa16@psulias.psu.edu 252 College & Research Libraries study, a prototype digital image library was developed and assessed. This pro- totype was to include a robust metadata schema appropriate to the content. This ef- fort was interesting, as the authors tried to reconcile some strikingly diverse metadata standards. Even more interesting were the challenges the authors faced in applying these schemas to the actual digital content selected for the prototype image library. Background and Description of the Visual Image User Study (VIUS) The provision of digital images for higher education has been an active area for research, design, and production projects in recent years. Many institu- tions have digitized collections and built institutional repositories, and continue to do so. Vendors are rapidly develop- ing so ware to assist them in those en- deavors. Products such as Luna’s Insight So ware, James Madison’s MDID, and OCLC’s contentDM enable the establish- ment of collections, addition of content, organization of images, and creation of metadata. Research projects during the 1990s, such as the Museum Educational Site Licensing Project (MESL) and the AMICO Library and Testbed, seemed to take a holistic approach, exploring production (multi-institutional coopera- tive production, in these cases), metadata coordination, system design, and user study (usually evaluating the results of the other efforts).1, 2 Research projects that followed these two important precedents have tended to focus on specific aspects of these problems. An NEH-sponsored grant project at the Rochester Institute of Technology’s Image Permanence Institute focused on the technical aspects of digiti- zation.3 The UCAI (Union Catalog of Art Images) project at UCSD has explored the problem of coordinating metadata stan- dards in a way that will facilitate shared cataloging of images.4,5 The Visual Image May 2004 User Study (VIUS) at Penn State was unusual among these research projects for taking a user-first approach, starting with a study of existing user habits and preferences to inform an approach to questions of metadata design, system features, and so on.6 Penn State University received a gen- erous grant from the Andrew W. Mellon Foundation in 2001 to conduct a 26-month project to assess the requirements for a digital image delivery system. The grant application grew out of discussions within the University Libraries about the scope, nature, and content of the digital library. The opportunity to examine one signifi- cant category of content—still images—as an aspect of digital library development immediately drew willing partners from outside the libraries, including museums, the Center for Education Technology Ser- vices, the School of Information Sciences and Technology, and faculty members in several departments. Together with the authors, they sought to understand how images are currently used on campus, what functions are supported by image use, who uses them, and the degree to which image use has become digital. The first phase of the VIUS focused on the study of user needs through focus groups, surveys, protocol testing, authen- tication log analysis, and other data collec- tion methods. The second phase involved the construction of two prototype image delivery systems: a centralized set of col- lections administered by the libraries and a peer-to-peer system supporting the shar- ing of digital objects in an authenticated environment (still in development). The final phase comprised the evaluation of these prototype services for usability, user satisfaction, and sustainability. The study yielded extremely useful data, which may be found in the final report on the VIUS Web site. Some of the key findings provide the background to The Academic Library as a Gateway to the Internet 253 the argument above about the design of digital libraries within an academic environment, as well as the assumptions that governed the design of a metadata infrastructure for the prototype database. Briefly, these findings are as follows: • There is considerable interest among faculty and students in using images for teaching and research (75% of faculty and 55% of students who responded to the survey currently use images). A consider- able number already maintain individual picture collections for professional reasons (44% of faculty and 44% of students say they have personal collections). A smaller, but significant, number are using digital images (50% of faculty and 45% of stu- dents), and many more expect to increase their use of digital images in the next three years (75% of both faculty and students). • Potential users of a digital image li- brary (faculty members, in particular) are more concerned about content-related is- sues than they are about retrieval-related issues; they are less concerned with how to discover images than with whether the image library will contain relevant images at all. Both their teaching and research needs tend to be highly individualized. • Many potential users are already building collections of images. The fact that they have specifically selected and retained these images to meet their re- quirements means that these collections are a critical component that needs to be incorporated within a digital library. • Faculty members are ambivalent about sharing their collections but are interested in the possibilities for collabo- ration and for shared development of core image collections. The study focused specifically on the arts, humanities, and environmental stud- ies for the purposes of the user study but anticipated that the production service to follow would contain collections and images across an even greater range of disciplines. The prototype centralized im- age library, on the other hand, was limited to art and architecture. Digital content was obtained from a variety of sources: commercial images, faculty collections, material from library book and archival collections, departmental and museum collections. One specific goal was to build a core collection of images to support the teaching of an introductory survey of landscape architecture. Metadata The function of descriptive metadata is to support discovery of appropriate resources. Because digital objects are cre- ated by specific user communities (and even individuals) with their own needs and vocabularies, the context in which a digital object originates has consider- able bearing on what type of descriptive metadata is appropriate to the object. The metadata about Notre Dame Cathedral necessarily would be different from that for a botanical specimen. Categories such as date, building name, and type of build- ing would be relevant for the former while the scientific name and taxonomy would be appropriate for the la er. Tony Gill has used the term “cultural infodiversity” to express the necessar- ily heterogeneous nature of cultural information, with the conclusion that “no single (meta)data schema fits all.”7 Although some descriptions are simple and generic, others are richer and domain specific. Access into these descriptions is often through the lowest common denominator provided by a crosswalk, ultimately leading the metadata editor into “mapping madness.”8 In the case of a digital library, with collections brought together from multiple disciplines and diverse sources, the digital images are both “infodiverse” and varied in terms of appropriate descriptive information. Clearly no single metadata standard 254 College & Research Libraries would be sufficient. The authors decided immediately to engage in their own ex- ercise in “mapping madness” in order to accommodate different descriptive metadata standards in a way that would support their interoperation. Given the focus of the study on the use of images in an academic environment, The authors quickly focused their a ention on three emerging descriptive metadata stan- dards: VRA Core, Dublin Core, and IMS Learning Resource Meta-data. VRA Core The focus on images and on the dis- ciplines of art and architecture for the prototype suggested that one standard that had to be accommodated was the Core Categories developed by the Visual Resources Association (VRA). The VRA Core consists of a single element set, with elements that may be repeated as many times as necessary to describe works of visual culture as well as the images that document those works. The distinction between the works and the visual images is a unique feature of the VRA Core. In an ideal system, a separate description would be created for each work (e.g., an individual building) and for each image of that work (e.g., a plan of the building or a photograph of the façade); each descrip- tion would include the element Record Type containing either the value “work” or the value “image.” The element set in- cludes elements specific to the discipline of art history, such as Style/Period. Quali- fiers also are suggested for some elements to provide more detailed semantics. Dublin Core Given that the potential digital library would need to encompass all disciplines, a common metadata language— Gill’s “lowest common denominator”—would most likely be needed. This suggested that the Dublin Core Metadata Element May 2004 Set (DC) should be examined. DC is a set of fi een descriptive elements designed to provide a simple means for describing resources to aid discovery and retrieval. Optionally, qualifiers may be used for a variety of purposes: to refine the semantics of an element (e.g., distinguishing a date of creation from a date of modification) or to identify the source of the data content (e.g., the use of the ISO standard for dates or of terms from a particular thesaurus). IMS Learning Resource Meta-data The authors’ partnership with the campus instructional technology center strongly suggested that the IEEE Learning Object Meta-data (LOM) 6-1 / IMS Meta-data Specification 1.2.1 (herea er referred to as IMS Learning Resource Meta-data) also be considered.9 Penn State is actively us- ing ANGEL, a learning management sys- tem that facilitates course instruction and enables reviews, quizzing, and chat (i.e., the delivery of various digital learning objects). Further, the Center for Educa- tional Technology Services is engaged in a variety of projects involving collaborative learning management supported by the IMS Learning Resource meta-data stan- dard.10 The purpose of the standard is to facilitate search, evaluation, acquisition, and use of learning objects, by learners or instructors, and to facilitate the shar- ing of learning objects by developing and exchanging catalogs and inventories of learning objects. The standard a empts to take into account the diversity of cultural and linguistic contexts in which learning objects and their metadata will be created and exploited.11 This emerging standard is not wholly new to libraries but has not yet been adopted widely or reckoned with in the design of digital libraries. The standard contains a number of unique features, perhaps most significantly its deeply hierarchical structure. It contains data elements appropriate to learning http:exploited.11 The Academic Library as a Gateway to the Internet 255 objects and their use. For example, the element LearningResourceType would be used for such objects as exercises, simula- tions, questionnaires, and exams. Metadata Structures: Challenges and Realities The basic assumption of the authors’ metadata design for the VIUS prototype database was that each of these three metadata standards needed to be sup- ported, that images included in the data- base might include metadata in any one or more of these standards, that the specific- ity of the metadata should be preserved in the VIUS database, and that some sort of mapping would be required to support searching across the entire database. Given these objectives, it was decided to create a “merged superset” of all the elements in the three standards. Because the standards have been created to meet the needs of specific communities, the authors wanted to preserve the context and specificity of the metadata created within each standard, with no loss of data in the process of merging. Thus, a schema was created that incor- porates all possible elements that might be carried by metadata contributed to the system. The schema provides an element that corresponds to each element in the three standards, allowing the specific- ity of any imported metadata to be pre- served. Where appropriate, the schema identifies equivalent elements across the three standards; these tend to correspond roughly to elements in the Dublin Core, and the hope was that these common elements would be sufficient to sup- port searching across the database. The schema, which could be expressed as an XML schema or DTD, would eventually support data transformations for import- ing and exporting data, for generating customized displays, and for creating a common query structure. The schema, in tabular form, is available on the VIUS Web site.12 The basic mapping was not terribly difficult. Many of the elements in each of the standards either had nearly equiva- lent definitions or were specific cases of elements in other standards. However, a closer examination of the mappings revealed difficulties. As the IMS “best practices” Web site states: “It is important to note that just because it is possible to map Dublin Core and IMS Learning Resource Meta-data elements to each other, this does not mean the elements are semantically or structurally equivalent. The reader should carefully study and un- derstand both the meaning and intended usage of each element before utilizing it in a meta-data record.”13 The authors found this warning to be amply warranted. There were semantic differences between the definitions of superficially similar data elements. Data elements using terms such as “type” or “format,” for example, were very likely to be intended to have quite different data values and might even have differ- ent functional uses. Another problematic set of elements are those that specify the “subject” coverage. Not surprisingly, the key elements of subject specification vary considerably from discipline to discipline. The VRA element Style/Period is a good example of a highly specific element; it is somehow related to the DC elements Subject or Coverage.Chronological but cannot be reduced to either DC element without significant loss of meaning.14 In addition to semantic problems, structural differences between the stan- dards presented a significant challenge. Moving between a profoundly hierarchi- cal standard such as the IMS Learning Resource Meta-data standard and a flat (albeit potentially qualified) set of ele- ments such as DC was difficult to model. For example, the straightforward DC http:meaning.14 256 College & Research Libraries element Subject is represented in IMS as Classificiation.TaxonPath.Taxon.Entry.15 Translating the two-level VRA structure (every element applies to either the work or the image) into standards that do not support this structure was impossible. The place of an element in a hierarchical structure serves to express some of its meaning. That context must not be lost when mapped to a different structure; it is unclear to what extent this is possible. The authors concluded that their merged superset was an important fea- ture of their design because it allowed the preservation of that context. It also meant that data transformations based on mapping between elements could be accomplished only at considerable risk of loss of meaning and should never be permanent transformations of the stored metadata. The critical question is whether the limited number of truly equivalent elements would be sufficient to support searching across the database. The mapping of data elements is com- paratively straightforward. However, the manipulation of data between elements by itself will not be sufficient to make the data truly interoperable. The main diffi- culties concern the meaning of the values contained in the elements. There are many sources for these difficulties. They may arise out of contextual differences in the use of language in different disciplines or differences in the role that the data element itself plays in imparting meaning to the values (the hierarchical context). Re- gardless of the source of the differences, mapping is about meaning. Anything that contributes to meaning in one context must be mapped into the new context or there will be loss of meaning. Unfortunately, it was not possible to test the authors’ metadata schema in the prototype. The need to select an available so ware platform for the prototype im- posed limitations on data structure, and May 2004 the schema could not be accommodated. However, the authors feel that their ap- proach was both valid and appropriate, and hope to test it in some other context. Metadata Values The difficulties that can arise in mappings between even the most carefully defined elements in published metadata specifica- tions have already been mentioned. Further complications arise given the variability of the values entered into elements. When speaking about values, the authors are talk- ing about the content within a descriptive element, the term entered therein. If the ele- ment is Date, for example, the string “2003” would be a value. However, several factors come into play with the assignment of values, for example, the ambiguity of lan- guage, the subjectivity of individual collec- tions, the subjectivity of persons describing the content, and the heterogeneous nature of disciplines and cultures that inform description. These factors are particularly significant in the university environment in which both faculty and students tend to work within particular disciplines, each with their own worldview, complete with vocabularies, taxonomies, and interests. The university, as a summation of these specific disciplines, must both nurture the specificity and support interdisciplinary serendipity. One word may mean different things. For instance, the word bridge may mean one thing to an engineer, another to a dentist, and quite another to a violinist. This reveals the relative sway that our disciplinary homes can have in informing our choices. Beyond this, just as the same words can mean different things, many words mean similar things. What one per- son calls the “bridge” between sections of a musical score, for example, another person might call a “transition.” Thus, the same object may be described in many ways. How is it possible to ensure that consistent http:Classificiation.TaxonPath.Taxon.Entry.15 The Academic Library as a Gateway to the Internet 257 terminology is employed for consistent values in describing an object? Further, as humans, our perspective and focus are unique. For example, a photograph taken in Venice might show a famous cathedral on a square with pi- geons, a canal, and several boats wending their way under a bridge. What “defines” this scene? What descriptive terms might it suggest, and how would they be as- signed? It seems clear that someone looking at the scene would identify the elements it contains in the same terms that they would use in retrieving the image. How a person chooses to retrieve the im- age will influence how he or she assigns descriptive terms to it. The architectural historian may well assign a “name” to the scene based on the cathedral’s name. The structural engineer might label the scene “bridge,” and the ornithologist might assign the name Columba fasciata. Clearly none of these terms, in and of themselves, defines the scene. At the same time, assignment of the label “Venice” does nothing to assist the searcher who is looking for pigeons. One can foresee the need for a system to permit the attachment of multiple, discipline-centric descriptive records to images. Such a system would permit one to choose to search the resource employ- ing the vocabulary of the art historian, for example, or the structural engineer, or the ornithologist.16 For now, problems with data values are best dealt with using authority control and content standards. In a published metadata specification, carefully defined elements are fixed and their design and structure are described in the standard’s Information Model. Accompanying the Information Model should be a Best Practices Guide, in which the designers of the standard a empt to guide real-world users in its proper application. Absent a specific set of rules or practices govern- ing the choice of values to be entered, or a controlled vocabulary from which to choose the terms (and a determination to employ the rules or the vocabulary or both), even the most rigorously defined elements cannot control the vagaries of humans and free association. To ensure interoperability and consistency among terms and elements, such problems are best countered by authority control and content standards. Functionally, this could be accom- plished using a toolkit that could be customized for a user ’s disciplinary worldview and would include guidance on selecting data elements and vocabulary. Such a toolkit could consist of Web-based forms, with blanks to be filled in and drop- down lists of suggested terms. This would encourage user contributions, preserve disciplinary understandings of terms, and maintain consistency among users. Contexts and Boundaries: Some Questions for the Future Experienced librarians and seasoned catalogers face numerous challenges in combining the worldviews of descrip- tive standards. What are the functional requirements for a system that would mediate these issues in a way that is simple, transparent, or invisible to the end user? In the end, the authors found their elaborate metadata schema of limited use. They discovered that available im- ages usually come with li le descriptive metadata. Faculty collectors might be able to supply a caption or title for each image (although sometimes this was a separate list with uncertain connection to the actual images). The authors were more surprised at the limited metadata that could be obtained from image vendors. This reality significantly shi ed the focus of the study. Instead of asking how to preserve complex metadata, the authors http:ornithologist.16 http:scene.At 258 College & Research Libraries were forced to ask how li le metadata would be required for discovery. This question is particularly important for image data. With text, inherited metadata could be mined, as could the content of the resource itself. With visual images, on the other hand, there is almost nothing in the resource itself that can be mined.17 One of the findings of the VIUS project is that individual collections are a signifi- cant component of a digital library and users therefore need to be able to add images dynamically, share images with colleagues, and merge their own images with those contributed to the database from other sources. This in turn requires that users be able to contribute metadata along with the images. In the authors’ plans, this would be accomplished techni- cally by providing a data input form with element names appropriate to the user’s disciplinary area. These elements could be mapped to appropriate elements in a generic interchange standard such as Dublin Core. The input forms for various disciplines could be prebuilt, with map- pings accomplished. Such a standard would itself have to be discipline agnostic, while providing a machine-understand- able means of conveying the contextual aspects of the record set needed to render its meaning accurately. A system such as this would have to deal properly with an aggregated set of records drawn from heterogeneous contexts. At the same time, it would have to enable interoperability among combined sets of records drawn from other similar (or sufficiently similar) contexts. The enabling lynchpin of such a system would be the establishment of a machine-readable standard for anchoring small text-based record sets containing vocabulary terms to subject- or context- specific controlled vocabularies. However, the addition of unmedi- ated metadata raises issues apart from the mechanical, technical aspects of the May 2004 transformations. Thus, again, how li le in- formation can be required without endan- gering the ability to retrieve the images? The requirements need to be kept modest if contributors are expected to provide the metadata. It also is important to consider how much oversight (i.e., manual editing) is needed to make up for either insufficient data or inconsistent data content. These issues need to be resolved if user-con- tributed metadata is to coexist with more complete, more consistent metadata. There have been notable efforts to pro- vide a technological framework in which to focus the still-inescapable human effort required to address these issues. The W3C Semantic Web exemplifies such efforts. W3C’s introductory explanation for its Semantic Web effort speaks to the need for “having data on the web defined and linked in a way that it can be used by ma- chines not just for display purposes, but for automation, integration and reuse of data across various applications.”18 It is useful to remember that although such linkages are maintained in a machine environment, their meaning relies on human a ention to their appropriate application. Another approach can be found in the IMS Vocabulary Definitions Exchange (VDEX) specification, in public dra as of this writing. VDEX “defines a grammar for the exchange of simple machine-read- able lists of values, or terms, together with information that may aid a human being in understanding the meaning or applicability of the various terms.”19 VDEX has potential applicability in the exchange of thesaurus fragments that could be unambiguously combined from multiple repositories. The contextual aspects thus captured could include the disciplinary context (by identifying the controlled vocabulary in use), as well as the specific thesaurus relationships among those terms expressed in machine- readable form. Here again, the technology http:mined.17 The Academic Library as a Gateway to the Internet 259 provides a framework for the interchange of values and their relation to each other, but the system supports, rather than sup- plants, the human effort. The discernment of subtle, but crucial, differences in mean- ing will likely forever elude brute force or the blunt instrument. Conclusion If a digital image library—or indeed any digital library collection—is to be suc- cessful within a university environment, it must accommodate the highly specific content that faculty and students are us- ing in their teaching and research. Such a digital library must be able to integrate content from various sources: existing library collections, individual faculty and student collections, content vendors, and other institutions. At the foundation of such a system must be a robust metadata schema capable of both supporting the needs of all the disciplines that comprise the university and interoperating with learning management systems through- out the university, transporting images and learning objects to the classroom, a dorm room, or a customized individual collection in response to user queries. The authors’ analysis of these metadata requirements caused them to construct a merged superset of several descriptive metadata standards. This was done in order to maintain disciplinary data ele- ments and vocabulary, while facilitating resource discovery across diverse collec- tions. The authors’ schema does allow for the preservation of metadata from each standard in all its specificity. The corresponding argument that it supports a sufficient level of interoperability has not been tested in application. However, similar schemas are being developed for other digital projects, and with their implementation will come real-world assessment. Over time, clarity should emerge regarding the issues raised by elements and their mapping in relation to context and its meaning for the values contained in those mapped elements. Another unavoidable issue concerns the limited metadata that generally comes with a digital object. Here a variety of solutions are needed. Active contributors of digital objects need to be provided with a toolkit for contributing metadata that is customized for their disciplinary world- view. Metadata obtained from outside sources (vendors, other institutions) may need to be edited manually. In both cases, authority control and content guidelines are needed to ensure consistency. University faculty and students are involved in the creation of content and are actively seeking additional content. A successful digital library program must integrate the content they create and provide with the highly individual- ized content that they seek. Metadata is the key to this endeavor. The institution must invest in the creation of appropri- ate standards and practices for acquiring and integrating content and its associated metadata; it must invest human resources in establishing these standards and prac- tices as well as in their consistent applica- tion. The guiding principles should be to preserve the disciplinary context of the digital objects, but also to allow them to be retrieved in a centralized system sup- porting a multidisciplinary universe. The authors have argued in favor of a merged superset of descriptive standards that would allow for maintaining disci- plinary relationships while facilitating cross-collection searching and discovery. They believe that their schema would allow for the preservation of discipline- specific descriptive metadata. Cognizant of the dynamic nature of the enterprise, the authors recognize that contributors to such an endeavor will need to adhere to modest metadata requirements, custom- ized for their disciplinary worldviews. 260 College & Research Libraries May 2004 Notes 1. The Museum Educational Site Licensing Project (MESL), an initiative of the Ge y Art History Information Project, was a two-year collaborative research effort among seven museums and seven universities to share digital images. 2. The AMICO Library and Testbed was a use and evaluation study of a collection of more than 20,000 documented, digitized images contributed by major art museums for licensed use by institutions of higher education. Available online from www.rlg.org/amicolib.html. [Accessed 30 January 2004]. 3. The report from 1999 identifies the key issues affecting image quality and explores ways to measure digital image quality. Available on the IPI Web site at h p://www.rit.edu/ipi. [Accessed 30 January 2004]. 4. UCAI (Union Catalog of Art Images) is available online from h p://gort.ucsd.edu/ucai/. [Accessed 30 January 2004]. 5. ARTstor, a digital repository of content in the history of art provided by various institu- tions, seems to draw from many of these research and production projects to provide a significant new digital image delivery service. Available online from h p://www.artstor.org. [Accessed 30 January 2004]. 6. The Visual Image User Study is available online from h p://www.libraries.psu.edu/vius/ reports.html. [Accessed 11 November 2003]. 7. Tony Gill, “Is That a Reference Model in Your Pocket? The CIDOC CRM & the IFLA FRBR,” PowerPoint presentation from Ready to Wear: Metadata Standards to Suit Your Project, An RLG- CIMI Forum, 12–13 May 2003. Available online from h p://www.rlg.org/events/metadata2003/gill. ppt. [Accessed 11 November 2003]. 8. Ibid. 9. In explanatory text intended to clarify the relation between the IEEE Learning Object Meta-Data (LOM) Dra Standard and IMS Learning Resource Meta-data, the IMS Global Learn- ing Consortium notes on its Web site;, “The IEEE LOM Dra Standard defines a set of meta-data elements that can be used to describe learning resources. This includes the element names, defini- tions, data types, and field lengths. The specification also defines a conceptual structure for the meta-data. The IEEE LOM Dra Standard is intended to support consistent definition of meta-data ele- ments across multiple implementations, but does not (at the time of this writing) include infor- mation on how to represent meta-data in a machine-readable format, necessary for exchanging meta-data. The IMS developed a representation of the meta-data in XML (eXensible Markup Language). The IMS Learning Resource Meta-Data Best Practice and Implementation Guide provides general guidance about how an application may use LOM meta-data elements. The IMS Learning Resource XML Binding specification provides a sample XML representation and XML control files (DTD, XSD) to assist developers with their meta-data implementations.” Available online from h p://www.imsglobal.org/. [Accessed 11 November 2003]. 10. Locally, CICERO, the Commi ee for Institutional Cooperation’s Educational Resource Repository project that enables the sharing of digital objects among members is another active digital project gaining momentum. Available online from h p://cicero.tlt.psu.edu/. Both CICERO and ANGEL support IMS. 11. View the home page of the IMS project at h p://www.imsproject.org/. [Accessed 11 No- vember 2003]. 12. VIUS data element definitions are available online at h p://www.libraries.psu.edu/vius/ reports/VIUSElementDescription.doc. [Accessed 11 November 2003]. 13. The IMS Learning Resource Meta-data Best Practices and Implementation Guide are avail- able online at h p://www.imsproject.org/metadata/mdbestv1p1.html. [Accessed 11 November 2003]. 14. VRA Core Categories, Version 3.0, a project of the Visual Resources Association Data Standards Commi ee, is available online from h p://www.vraweb.org/vracore3.htm. [Ac- cessed 11 November 2003]. From the VRA Core: STYLE/ PERIOD Qualifiers: 1 Style/Period.Style 1 Style/Period.Period 1 Style/Period.Group 1 Style/Period.School 1 Style/Period.Dynasty 1 Style/Period.Movement 1 http:h�p://www.imsproject.org http:h�p://cicero.tlt.psu.edu http:h�p://www.imsglobal.org http:h�p://www.artstor.org www.rlg.org/amicolib.html The Academic Library as a Gateway to the Internet 261 Description: A defined style, historical period, group, school, dynasty, movement, etc., whose characteristics are represented in the Work or Image. Dublin Core: COVERAGE, SUBJECT 15. The hierarchy can best be illustrated by an example expressed in SGML-style tags. The subject term “Landscape Architecture” would be expressed in IMS as Discipline LCSH Landscape Architecture This identifies landscape architecture as a discipline (subject) taken from the Library of Con- gress subject headings (LCSH) list. 16. One could foresee a useful purpose in permi ing students in a class section to associate their own descriptive terms with a learning object accessed through a course management system. Such user-supplied metadata could be persistent through a semester and then lapse or remain confined to that section’s view of the resources (perhaps even student by student). Such an ap- plication of user-supplied metadata approaches the concepts captured in the term “democratic indexing,” as described in Brown et al, “The Democratic Indexing of Images,” New Review of Hypermedia and Multimedia (2): 107–19. 17. Most research on mining image data deals with pa ern recognition and provides nothing in the way of content analysis. 18. See W3C’s Semantic Web site at http://www.w3.org/2001/sw/. [Accessed 8 October 2003]. 19. IMS Global Learning Consortium’s VDEX site is located online at h p://www.imsglobal. org/vdex/index.cfm. [Accessed 8 October 2003]. http://www.w3.org/2001/sw