| ||
 | D-Lib Magazine |
| Henry Pisciotta |
| AbstractPenn State University is conducting an extensive and systematic assessment of its needs for digital image delivery. This work will include the development of a prototype image delivery system. Though focused on a local solution, the Visual Image User Study (VIUS) is seen as contributing to national efforts to develop digital picture libraries to serve the many new uses for digital images for teaching and research. A rigorously client-centered approach is expected to yield new data that will help identify priorities among the many issues in image retrieval. This article focuses on questions addressing image delivery that must be answered; lists the purposes of the VIUS needs assessment, and discusses ways of coping with the shortcomings that can be typical of needs assessment. Issues in Image DeliveryPictures have the power to convey enormous amounts of information within the precious space of a computer screen and are expected to become an important part of the digital library and learning environment. But compared to the large number of sources and systems for electronic texts, the delivery of digital pictures seems underdeveloped. The chief reasons may be historical. Electronic text files have been practical for a much longer time than pictorial data. Even in non-electronic forms, publishers of pictures have been fewer than publishers of texts. Shared cataloging and a broad selection of indexing services are long-established features of providing access to texts but barely exist for pictures. Most of the work on the description and organization of picture collections has taken place within rather narrow disciplinary boundaries. Slide and photograph collections at most universities are highly decentralized -- operated by separate departments or even individual faculty. Although this history has produced plenty of fine slide and photograph collections, many people have recognized the potential of digital images1 to alter this path of development, since digital pictures are easily shared and are now of sufficient quality to support a variety of uses. But how should that path be altered? The following is a selection from the many questions regarding the future of digital image delivery, as chosen from a review of recent literature and development projects. Are all picture collections now interdisciplinary?The Museum Educational Site License Project (MESL) was an important point of departure for studying the use of digital images in academic settings. As part of the analysis of that project, which distributed art museum images to several universities, researchers predicted that users of networked digital images would include people from a wider variety of disciplines than has been typical of slides and photographs.2 If this proves true, what are the implications for planning a retrieval system? In her excellent review of the development of cultural heritage image databases, Christie Stephenson has observed that "Moving from the analog to the digital world, it becomes increasingly difficult to characterize the users of our delivery systems." She emphasizes the traditional approach of focusing on the needs of clearly defined user groups in order to understand system requirements, but recognizes the increasing difficulty of doing so, and the possibility of an alternative: "Or are we building hybrid systems, merging collections created for some or all of these purposes, into large supersets, where audience and aims become increasingly indeterminate?"3 So far, in the academic world, the answer to her question is usually "No." Yet among the wildly popular media search engines available on the web, the answer seems to be an emphatic "Yes!"4 How can we measure the significance of this remote, indeterminate audience? How much can access be federated?To some extent, the decentralized history of picture collections can be seen as a potential strength. Distributed computing systems are well suited to multiple sources of data production and sharing of complex files with only a few key elements of standardization. In the race to reach the much-discussed critical mass needed for a digital picture collection to prove useful, having multiple producers should be an asset. Non-profit projects such as the AMICO Library5, OhioLINK6, and now ArtSTOR7, are making exciting progress in the direction of large-scale image systems with international service potential. A variety of equally useful commercial sources have appeared in recent years.8 Some free sites are also supplying important images of good quality to any users.9 Yet local needs persist. Perhaps one reason that some local production of digitized images will always take place is that there are no limits to the imagination of a good teacher. In fact, as the sources for digital images increase - some homemade, some licensed, some free -- methods of helping students and teachers make sense of them will become increasingly important. Can using many sources be made as convenient as using one? Some institutions are attempting to federate access to images through several means: designing a single database for the management of several collections; creating a retrieval database which is separate from, but fed by, several collection management databases; or employing search agents to query several databases.10 Along with the many new sources for digital images have come arrays of rights and permissions. Can federated access help to manage the many privileges that might be associated with an image or with a user? Should retrieval be semantic or visual?Pictures may be described or they may be seen. The efficiency of seeing a picture is acknowledged in most interface designs, at least by providing screens filled with thumbnail images. But getting to that display usually requires selecting words that match to cataloging terms. This matching process is difficult for any type of material but doubly so for pictures because of the need to translate from visual to verbal expression. How should pictures be described? In some disciplines there are many fine tools to help with this description, but their application varies widely from one collection to another.11 The difficulty of description is increased again with digital picture collections if we expect interdisciplinary use. Some have suggested that user-contributed descriptions could be a useful way to represent multiple points of view.12 Other researchers hope to circumvent the need for descriptions by visual approaches to indexing. Visual indexing (sometimes called "content-based indexing") strategies include matching by colors, shapes, textures, or other purely pictorial criteria. So far, these methods have proven most effective in limited problem domains (mug shots and fingerprints, for example). Hybrid systems combine visual indexing with semantic indexing.13 Even if the visual component of a retrieval system is limited entirely to thumbnail displays, every interface strives for the right blend of searching and browsing. What is the new relationship between retrieval and use?With a traditional card catalog, a line could easily be drawn between retrieval and use. "Retrieval" was finding a catalog record and a place on the shelf. "Use" was opening a book and reading. Now that retrieval and use both take place in the microcomputer, the line is not as clear. This obvious situation has some special implications for image databases. Users may want "light table" displays to arrange lectures, or comparison functions for classroom presentation. They may want to use the retrieval system for projected display or to permanently store sequences of images as reusable presentations. Outputting images and descriptions in formats compatible with other presentation software or course delivery software may be desirable. As courseware delivery systems become more useful, instructors seem to want their virtual students to have access to large amounts of information as part of their digital learning environment. Features of course delivery software and information retrieval software are beginning to overlap. Image delivery is becoming an important common ground in these types of software developments. Image manipulations (zooming, cropping, adding annotations, etc.) might be needed. Perhaps some of the descriptive data should remain associated with the image file during these uses. What use tools need to be part of a retrieval system? Should there be one interface or several?Other specialized tools may appeal to certain categories of users - for example, mapping displays for global positioning data, three-dimensional displays and rotations, etc. As the features of a retrieval system multiply, user-controlled customization has been suggested as a way of preventing flexibility from confounding usability. Are these the right questions?These questions only partially represent the issues that can be found in the literature. Issues related to metadata, image quality, sustainable system design, and many others could be added. Yet even this selection is too long and wide-ranging to be addressed by a single development project. For this reason, Penn State has chosen to identify priorities through rigorous needs assessment -- a study of our users' perceived needs and current behaviors, as well as a review of the existing assets that can be employed to serve those needs. We hope to identify the questions of critical importance. Needs AssessmentThe purpose of the VIUS needs assessment is to:
The primary research questions are:
These are local questions, but to answer them well will require addressing the types of questions that introduce this paper. Topics examined will include the content, management and use of existing image collections (analogue and digital); users' hopes for future content and uses; the perceived value of various retrieval, manipulation, and display tools; identification and coordination of institutional stakeholders in image delivery; and the utility of existing software tools and metadata standards. To construct answers to our research questions, we intend to use a rich combination of assessment techniques. The list of techniques has been selected to overcome the weaknesses of individual assessment tools, particularly as regards combining user perception measures with measures of actual behaviors, qualitative with quantitative measures, etc. Assessment techniques we will use include: Information retrieval measures
User perception measures
Other measures
In undertaking needs assessment, we recognize three potential pitfalls:
AudienceThe broadly interdisciplinary Penn State project focuses on the arts, humanities, and the environmental sciences. These were chosen because we know that images are available to serve them and because Penn State's students and teachers in these fields are active users of images. They also represent a challenging array of approaches to teaching and learning. In order to characterize this audience in terms that will translate into system functions we seek to identify the needs of three types of system users:
Good solutions will require an excellent understanding of the needs of each of these types of users. These roles are seen as overlapping. Active learning techniques sometimes turn learners into teachers. Both may be involved with managing collections or subsets of collections in their work. By carefully studying the image needs of these three educational roles, Penn State hopes to boost understanding of this aspect of digital library development. Assessing New TechnologiesSince image databases are still a new and, to many users, unexpected aspect of electronic information services, Penn State looks to the model of new product planning and development, where needs assessment has been shown to be a critical aspect of technology-driven innovation. As part of the mix of assessment techniques, a basic procedure of new product development will be employed: a series of consultations with users, progressing from general issues, by several stages, to potential solutions and their specific implementations.15 Prototype Development and EvaluationA danger of needs assessment is that its results may remain untested and unemployed. To reduce that risk, Penn State will develop a prototype image delivery system. Determination of the features and content of this prototype will be based upon the conclusions of the needs assessment. The critical issues which surface during needs assessment should be explored further through evaluation of the prototype. The prototype content must also coordinate with actual teaching activities. We estimate that the prototype should provide access to 3,000 to 8,000 locally mounted images (and possibly additional external sources). The number of images and the level of description provided in the prototype could vary widely with the results of needs assessment and the specific requirements of the teaching activities selected. The system will need to facilitate many forms of learning (classroom presentations, distance education packages, group assignment, independent learning, etc.) A new, more specific set of research questions will be generated for this phase, but their most general statement will be: 1) In what ways does the prototype meet and/or fail to meet the requirements and priorities identified in the needs assessment? and 2) Do user interactions with the prototype reveal additional needs assessment information? For this evaluation, we expect reliability testing, protocol analysis, focus groups, and intensive interviews to be important tools. New software development will be kept at a minimum. In preparation for prototype development, the major software applications relevant to image delivery will be reviewed and selected. If existing tools do not include features of importance to users, these will be developed in-house, if feasible. Participation in this project by personnel of Penn State's Educational Services Technology will help ensure that image delivery is coordinated closely with local and international courseware delivery systems and standards. Further information about the VIUS project is available at: <http://www.libraries.psu.edu/crsweb/vius/index.html>. AcknowledgmentsPenn State is grateful for the Andrew W. Mellon Foundation's generous support of this project. The authors listed here are those primarily responsible for the project proposal. They wish to thank many people at Penn State who have made substantial contributions to this plan, particularly: Nancy Eaton, Dean of the University Libraries; John Harwood, Director of the Center for Education Technology Services; Sally Kalin, Associate Dean of University Park Libraries; and Amanda Maple, Head of the Arts and Humanities Library. Recent additions to the project team include Michael Halm, from the Center for Education Technology Services; Joni Barnoff, from Library Computing Services; and John Attig and Ann Copeland from the University Libraries. Notes and References[1] In this paper, the term "digital image" is used to describe still pictorial data files rather than image files that primarily convey text. The proposed project focuses upon collections of still images in the major image file formats (JPEG, GIF, TIFF, etc.). Images of text pages (scanned manuscripts or books) are excluded. [2] Besser and Yamashita, "The Cost of Digital", "Executive Summary", 8; and Lack, R., "The Patterns of Slide Library Circulation: a Study", 4-5 in: Howard Besser and Robert Yamashita, The Cost of Digital Image Distribution: The Social and Economic Implications of the Production, Distribution and Usage of Image Data, Report produced by the School of Information Management & Systems, University of California, Berkeley (July, 1998) <http://sunsite.berkeley.edu/Imaging/Databases/1998mellon>. Data from a subsequent study supports that prediction: Henry Pisciotta and Tracey DePelligrin, "Demographics of AMICO Library Users at Carnegie Mellon University", (in preparation). [3] Christie Stephenson, "Recent Developments in Cultural Heritage Databases: Directions for User-Centered Design", Library Trends, 48 (Fall, 1999) pp. 422-424. [4] On the popularity of web search engines for retrieval of pictures see, for example: Abby Goodrum and Amanda Spink, "Image Searching on the Excite Web Search Engine", Information Processing and Management: An International Journal, 37, #2 (2001) p. 295-311. [5] AMICO Library (http://www.amico.org/home.html) is a subscription-based digital image library of over 50,000 art images for use in education, research, and teaching. The Art Museum Image Consortium is a not-for-profit association of art-collecting institutions working together to enable educational use of their digital documentation. [6] OhioLINK (http://www.ohiolink.edu) is the Ohio Library and Information Network, a consortium of Ohio's college and university libraries and the State Library of Ohio. [7] The ArtSTOR project is described briefly in: Florence Olson, "Foundation Plans Digital Archives for Art and Architecture", Chronicle of Higher Education, (April 20, 2001) p. A54, or in Andrew W. Mellon Foundation, 1999 President's Report, <http://www.mellon.org/awmar.html>. [8] The AP Photo Archives (http://ap.accuweather.com/apphoto/index2.htm), Corbis (http://corbis.com), and Saskia, Inc (http://www.saskia.com) are examples of commercial enterprises with digital image databases available through license agreements. [9] For example, the Society of Architectural Historians (http://www.sah.org/imagex.html) has created a digital image exchange program in which members contribute their own slides of buildings. This project posts images of objects mentioned in current textbooks about the history of architecture and provides free access to the images to members and students over the Web. [10] Examples of these approaches include the Visual Image Access system at Harvard (http://hul.harvard.edu/ois/systems/via/index.html), and the Image Services system of the Digital Library Production Service at the University of Michigan (http://images.umdl.umich.edu/i/image/all). [11] For example, within art and architecture the Getty Vocabularies Program has produced structured vocabularies for artists names, geographical names, and art and architecture subjects (http://www.getty.edu/research/tools/vocabulary) and the Visual Resources Association has developed the VRA Core Categories to guide the description of images (http://www.gsd.harvard.edu/%7Estaffaw3/vra/vracore3.htm). [12] For example: Brian C. O'Connor, Mary K. O'Conner, and June M. Abbas, "User Reactions as Access Mechanism: An Exploration Based Upon Captions for Images", Journal of the American Society for Information Science, 50, #8 (June, 1999) pp. 681-697. [13] A variety of visual indexing techniques are reviewed in several articles within: Beth Sandore (ed.), "Progress in Visual Information Access and Retrieval" (Thematic Issue), Library Trends, 48, #2 (Fall, 1999). Hybrid systems include MetaSEEK (http://ana.ctr.columbia.edu/metaseek), SIMPLIcity (http://wang.ist.psu.edu/IMAGE), and others. [14] On customization of interfaces, see for example a review of the recent work with library web pages: Eric Lease Morgan (ed.), "Special Issue: User-Customizable Library Portals", Information Technology and Libraries, 19, #4 (December, 2000). [15] C. Merle Crawford, New Products Management, Burr Ridge: Irwin, 1994, and John A. Hall, Bringing New Products to Market, New York: AMACOM, 1991. Copyright 2001 Henry Pisciotta, Roger Brisson, Eric Ferrin, Michael Dooris, and Amanda Spink |
| |
| Top | Contents |
| |
| D-Lib Magazine Access Terms and Conditions DOI: 10.1045/july2001-pisciotta
|
