Image analysis and indexing in North America: a survey

Art Libraries Journal Summer 1982 51 Image analysis and indexing in North America: a survey Thomas H. Ohlgren, Those in charge of image collections in North America have found it difficult to agree upon standards for classifying and cataloguing their material, and many individual and incompatible systems have been developed. However there has been some recent progress in co-ordination, and this paper surveys some inter-institutional projects attempting to create standards, based on an awareness of user needs, primarily the Image Access Society, the Library of Congress list of subject headings for the Prints and Photographs Division, the Picture Division Thesaurus of the Public Archives of Canada and the Art and Architecture Thesaurus. This paper is a continuation of a survey, "Subject indexing of visual resources: a survey", published in Visual Resources: an International Journal of Documentation, Vol. 1, no. 1, Spring 1980, pp.67-73. As David Vance points out in "Computers and the Fine Arts in the United States," museums in the United States do not operate under a single governmental authority/ 1 ) Consequently, each institution has developed highly individualised, if not idiosyncratic, approaches to the management of data about its collection. Due to dozens of different cataloguing and indexing schemes, communication, co-operation, and especially co-ordination of efforts are virtually impossible. The same problem plagues photographic, slide, and film collections. As I have stated elsewhere, archivists and curators have not been able to agree upon standards for classifying and describing art works or their photographic surrogates. Instead of seeking a co-ordinated approach, archivists have erected mutually unintelligible Towers of Babel/ 2 ) And speaking of Babel, among those collections utilising computers to manage their collections, there is also a proliferation of different computer software programmes: GRIPHOS, SELGEM, STAIRS, TAXIR, GIPSY, BIRS, MARK IV, and INQUIRE, to name only a few. It goes without saying that the machine-readable records of one collection cannot easily be merged with the data of another collection utilising a different software system. Thus, networking, or pooling of data, is not 52 Art Libraries Journal Summer 1982 possible. In short, the use of both independently-created standards for describing visual images and incompatible software programmes for processing the data, have resulted in chaos for the researcher desiring to locate visual documents illustrating a common subject or theme, in a number of different collections. In spite of these seemingly insurmountable difficulties, some progress is being made in North America to face the issues of the lack of subject standards, the lack of institutional co-operation, and the role of computers in providing subject access to visual resources. What follows is a brief survey of those inter-institutional projects designed to address the subject needs of researchers at large. Although much could be said about individual institutional projects in North America, such as those at the Yale Center for british Art/ 3 ) the National Museum of American Art (Smithsonian)/4) and the Medieval Photographic Archive (Purdue)/ 5 ) all of which are using computer-aided approaches. I want to highlight those activities directed specifically at inter-institutional co-operation and co-ordination. Image Access Society of North America The Image Access Society was founded in August 1979 when a group of archivists, curators, art historians, museologists, slide and photograph librarians, and information specialists, met at the first International Conference on Computers and the Humanities, held at Dartmouth College in Hanover, New Hampshire. In addition to hearing seven papers on mainly institutional projects/ 6 ) the group agreed to pursue collectively the following goals: 1) to survey the existing iconographic classification schemes in use and to compile a collection of this material; 2) to investigate the establishment of a common critical nomenclature for describing and evaluating the various iconographic schemes (e.g., analyse each scheme in terms of precision, specificity, exhaustivity, and so forth); 3) to survey by questionnaire the various constitutent members of the Society to gather information on the various approaches to subject access in North America; 4) to sponsor or to encourage pilot projects in the areas of thesaurus construction and subject indexing; 5) to recommend, after due consideration, a common set of standards for iconographic classification and description; 6) to organise a conference to discuss further how the goals are to be realised. Art Libraries Journal Summer 1982 53 The next meeting of the Image Access Society was held on August 1-3, 1980 at the Belmont Conference Center in Elkridge, Maryland. The meeting was organised and chaired by Philip Leslie, Registrar of the Smithsonian Institution, who explained the need for the conference as follows: "People who collect works of art, whether originals or reproductions in various media, often need to retrieve information by what can be termed the subjects of the works. With reference to a work of art, a 'subject' might be something like a pictorially implied theme, a type of person illustrated, a type of object illustrated, a location, etc. The need is very similar to the need to locate written material by subject. As with a written work, facilitating retrieval involves assigning subject terms to each item.and creating some sort of file through which works can be identified by these subject terms. Whether the file is a manual one or a computerised one is not particularly relevant. What is relevant is the philosophy behind the structuring of the index t e r m s . . . There is now little doubt that awareness of information needs is growing rapidly in the entire collections community. A number of attempts have already been made to develop subject lists, some of them promising and some of them bad enough to make an experienced vocabulary specialist shudder. During the last few years, a number of people interested in visual resources have been watching the trend with growing concern. They think people need some help in learning how to develop sound indexing vocabularies for their own collections. They see enough commonality among art-related collections to suggest the possibility that standardised lists might be applied to different collections. And they are afraid that any chances there are for compatibility, or even mutual help with local projects, are apt to diminish in proportion to the number of dissimilar lists being developed". The main purpose of the Belmont meeting was to investigate the desirability and feasibility of developing uniform standards and logic for the subject indexing of art works or their reproductions. To facilitate discussion at the conference (no formal papers were presented), a questionnaire was sent to those conference participants who indicated they would attend. The questions were as follows: 1) What types of visual resources are in your collection? 2) What is the size and scope of your collection? 3) What is the growth of your collection? 4) What avenues of access do you provide? 5) Do you provide subject access? If so, what specific sources (authority 54 Art Libraries Journal Summer 1982 lists, thesauri) do you use? 6) Is the subject source "in-house" or "out-of-house"? 7) If "out-of-house," how many other collections adhere to these standards? 8) Are you satisfied with the subject standards you are using? 9) Ideally, what kinds of standards and/or thesauri would you like to see developed? The twenty-four participants, who included museum administrators and professionals, photo-archivists, art historians, slide and photo librarians, cultural historians, and information specialists, represented collections ranging from original art works (prints, drawings and paintings), museum artifacts, and maps, to photographs, slides and films. Although space does not allow for a detailed summary of the meeti n g , ^ which was more of a retreat with some thirty hours of debate and discussion, I will try to summarise the major issues and questions raised. 1) Is there a need for "universal" descriptive standards, covering the universe of human knowledge, to aid the archivist in the assignment of subject descriptors to his or her collection? If so, what are the major subject categories and how should they logically be structured? 2) Assuming that universal descriptive standards were created, would these standards, in the form of controlled vocabularies, authority lists, and thesauri, be relevant to the specialised needs of individual collections? That is, would a broad scheme be specific enough to anticipate the subject needs and interests of researchers in specialised fields? 3) What is the difference between "free language indexing" and "controlled language indexing"? Christopher Seifried (National Photography Collection, Public Archives of Canada) defined the terms as follows: "In free language indexing the person responsible does not assign any terms to the document, but uses the terms which are already attached to the document, as, for example, the title of the material, or key terms identified in an abstract." By contrast, "in a controlled system the indexer imposes terms, and eventually these terms are collected in an authorised list which will guarantee consistency in vocabulary." 4) In controlled vocabulary indexing, what are the various ways of structuring the terms? Philip Leslie identified and defined five different logics: a) uncoded monohierarchical classification; b) coded monochierarchical classification; c) library-type subject headings; d) conventional thesauri with cross-references; and e) permuted and co-ordinate indexes. 5) What is a thesaurus? Robert Chenhall, Director of the New Mexico Museum of Natural History, offered the following: "A thesaurus is any kind of structured authority file; it can be hierarchical or alphabetical." Art Libraries Journal Summer 1982 55 Philip Leslie added the following stipulation: "An authority list per se normally consists in its root form of nothing but the terms that you are authorised to use in indexing; it is just a straight alphabetical list of what is allowed. If you begin to show linkages among the terms, listing some that cannot be used and telling what to use instead (see references), or if you begin to put 'see also references' in, or if you expand the 'see also references' into three kinds — broader generically, narrower generically, or related in some way - then you have a thesaurus." 6) What would be the minimal standards for a practical system of subject access? After much discussion, the group arrived at the following basic requirements: a) The system should be universal and interdisciplinary. b) It should be consistently applicable from one institution to another and within departments of the same institution. It should be equally intelligible and useful to the various indexers within the same department. c) It must be applicable to various levels of complexity. d) It should be portable or distributable by some means. e) The indexing vocabulary used should elicit approximately the same visual images to people of different backgrounds. f) It should be possible for anyone to get access to the system and to reach comparable answers with comparable questions. g) It should be open-ended or expandable to allow for the modification of existing terms or the addition of new ones. In addition to the discussion of broad theoretical issues, specific thesaurus construction projects were evaluated. Those projects of direct relevance to the problem of standards for the subject indexing of visual resources include: 1) the Library of Congress Subject Headings; 2) the Picture Division Thesaurus of Iconographic Terms (Public Archives of Canada); and 3) the Art and Architecture Thesaurus. List of Subject Headings for the Library of Congress Prints and Photographs Division Although her project has not been officially endorsed by the Library of Congress, Elisabeth W. Betz, Picture Cataloguing Specialist in the Prints and Photographs Division, has recently completed a preliminary, alphabetically-arranged list of 3,500 subject headings with cross references, which will be used eventually to increase subject access to the ten million prints and photographs in that division. According to Betz, each of the 65,000 entries in the previously-existing Divisional subject index was compared with the subject headings in the Library of Congress Subject A rt Libraries Journal Summer 1982 56 Headings (8th. ed.) and was modified as needed to apply to visual images. While Beta's list was specifically created for the collections of the Prints and Photographs Division, it offers many subject headings that might be applied to general picture collections in museums, archives and historical libraries. A copy of the list may be obtained from the Library of Congress Cataloguing Distribution Service for £15. Picture Division Thesaurus (Public Archives of Canada) The Picture Division, in co-operation with the National Inventory Programme, is actively developing a computer-stored thesaurus of iconographic terms. The thesaurus will provide a controlled indexing vocabulary for indexers in some 150 Canadian federal, provincial and municipal museums. In addition, the computerised inventory programme will permit on-line access to data about some 100,000 works of art in the Picture Division, which is tne country s largest repository of early Canadian iconography, consisting of watercolours, engravings, etchings, lithographs and prints. The thesaurus will encompass the following major categories: architecture, costume,.activities, artifacts, flora, fauna, insignia, landscapes, people and transportation. According to Denis Castonquay, the thesaurus is a structured vocabulary using three levels of specificity: 1) a controlled list of broad classification categories (e.g., the major categories listed above). 2) a controlled list of classification terms, consisting of carefully-defined subdivisions of the broad categories. 3) an open-ended, expandable list of specific indexing terms. ARCHITECTURE DOMESTIC ARCHITECTURE APARTMENT HOUSES CABINS COTTAGES FARMHOUSES GATEHOUSES HISTORIC HOUSES IGLOOS MANSIONS ROW HOUSES TENTS TEPEES WIGWAMS Art Libraries Journal Summer 1982 57 The indexing terms will be accompanied by cross references to broader, narrower and related terms. Rules for constructing the thesaurus are based upon the American National Standard Institute's manual, Thesaurus Structure, Construction, and Use. (ANSI Z39) In addition to developing iconographic standards, the Picture Division has formulated standards for the descriptive cataloguing of its holdings. Douglas E. Schoenhen defines descriptive cataloguing as "the recording of all relevant information pertaining to its unique identity within the collection, the artist who created it, its physical characteristics, and its history up to the present.''^8) This information, Schoenherr states "was crucial in authenticating and establishing the aesthetic, iconographic and historical significance of a given work." Instead of using existing records, which varied greatly in consistency and quality, the decision was made in the Picture Division "to create new records directly from the works in our collection according to high and uniform standards of description, and to create these new records specifically for the computer." The standards consist of 68 fields of data, which divide into four main sections: archival, artist, physical and historical data. The Canadian project, in my opinion, provides a model for other institutions to emulate. Although the vocabulary developed for specifically Canadian iconography would not be relevant to collections outside of Canada, the logical structure of the thesaurus, the scientific approach to the creation of the iconographic vocabulary, and the state-of-the-art computer component, certainly make this one of the major subject indexing projects in the world. Art and Architecture Thesaurus (AAT) In direct response to the lack of a standard vocabulary for subject indexing, Pat Molholt (Rensselaer Polytechnic Institute, Troy, New York 12181), Toni Petersen (Librarian, Bennington College), and Dora Crouch (also Rensselaer Polytechnic) have recently announced a project to create a thesaurus for use in art and architecture.^ The proposed thesaurus will function both as an authority list for the selection of descriptors by indexers and as a guide to point to information or objects which are most relevant to the searcher's needs. The AAT group has received two planning grants. One grant, from the Council on Library Resources, is enabling the group to survey twelve current indexing and thesaurus projects and to write a report on their findings (available for S3.00 from Pat Molholt). Another grant, this time from the National Endowment for the Humanities, will fund the development of a standard vocabulary for the field of architecture. The plan of work for the architecture thesaurus consists of com- 58 Art Libraries Journal Summer 1982 piling the existing lists of architectural terms from a number of sources (Library of Congress subject headings, 9th. ed., t\it Architecture thesaurus of the Public Archives of Canada, the thesaurus of the Journal of the Society of Architectural Historians, the subject lists provided by the International repertory of the literature of art (RILA), and the Avery index to architectural periodicals) and analysing the vocabularies for form, structure, style and contents. The next step will be to structure the subject headings chosen for inclusion into a hierarchy and to code the results for computer storage, retrieval and analysis. The final stage, yet to be funded, will be to create a similar thesaurus for the fine arts. It is much too early to tell if this ambitious project will provide the subject standards we are all seeking. Given its total dependency on governmental support, it would be a miracle if the project received the $500,000 needed to complete the proposed thesauri. Finally, we must seriously ponder the cautionary note sounded by Philip Leslie in his concluding remarks at the Belmont Conference: "I came here convinced of something different from what a lot of others believed. There were a number of people who thought that out . of this conference we would get the wherewithal to develop a universal thesaurus for subject indexing of visual resources of all types and all fields. I tried very hard not to comment, because I felt that this goal was impractical and illusive. I feel that the easiest way to generate a thesaurus is to allow the contents of a specific collection to determine what was to be indexed and therefore what concepts and objects need to be covered in your subject descriptors. I feel that if you have several collections covering a specific subject area, such as medieval iconography, then it is possible to develop vocabulary, authority lists and thesauri that cover specific areas. I think that vocabulary controls are best built-up like building walls out of bricks. You start at the collection level, and combine collection vocabulary lists to form subject area vocabulary lists. Whether you can reach the level of combining lists to form a megalo-classification has yet to be determined. I have my doubts. But I do not feel that is the place to start. I do not think you start at the top; in this case you start at the bottom." Conclusion It is clear, then, that the major issue being discussed in North America is standards, specifically standards for the classification and description of visual images. This is, of course, ironic because many institutions, museums, and archives already have well-developed database management systems for their collections. It seems that the siren of technology lured Art Libraries Journal Summer 1982 59 many archivists away from the fundamental question of uniform standards for classifying images to the lusty pursuit of the most complex, sophisticated and expensive computer systems. Although a system may work well for a specific collection, it is so individualised that the pooling of the stored data with similar data in another collection is impossible. Even if uniform standards were at long last to be agreed upon, it seems unlikely that those archives already committed to an "in house" approach could afford to re-describe their iconographic information. The real hope is that curators of collections, who are contemplating the conversion of their manually-produced records to machine-readable form, will not repeat the mistakes of the past. Thomas H. Ohlgren Medieval Photographic A rchive Purdue University West Lafayette, Indiana References (1) VEZINA, Raymond (Ed.). Computerised Inventory Standards for Works of Art. Montreal: Fides, 1981,pp.89-97. (2) OHLGREN, Thomas H. Subject Indexing of Art Works, pp.73-77 in Computerised Inventory Standards for Works of Art. Montreal: Fides, 1981, Ed. Raymond V6zina. (3) SOBINSKI-SMITH, Mary Jane. The Yale Center for British Art: The Photographic Archive and iconographic access. Visual Resources, Vol. 1, no. 2/3, Fall 1980/Winter 1981, pp.173-87. (4) FINK, Eleanor. Subject Access to the Photographic Reproductions of American Paintings at the National Collection of Fine Arts, pp.229232 in Data Bases in the Humanities and Social Sciences. Amsterdam: North-Holland Publishing Company, 1980, Eds. Joseph Raben and Gregory Marks. (5) OHLGREN, Thomas H. Computer Indexing of Illuminated Manuscripts for Use in Medieval Studies. Computers and the Humanities, Vol. 12, no. 1/2, 1975, pp.189-199; Subject Access to Iconographic Data Bases: Theory and Practice, pp.245-250 in Data Bases in the Humanities and Social Sciences, op.cit. (6) All of the papers have been published in Data Bases in the Humanities and the Social Sciences, op.cit. 60 Art Libraries Journal Summer 1982 (7) RODDY, Kevin. The Belmont Conference on subject access. Visual resources: an international journal of documentation, Vol. 2, nos. 1/2/3, Fall/Winter 1981/Spring 1982, pp.101-111. (8) Standards for the Descriptive Cataloguing of Paintings, Drawings, Watercolours, Prints and Posters In the Picture Division (Revised edition of 1979). (9) MOLHOLT, Pat. The Art and Architecture Thesaurus: A Project Report. Visual Resources, Vol. 1, no. 2/3, Fall 1980/Winter 1981, pp.193-199. Discussion notes Points arose concerning the Belmont Conference. Criticism was voiced that only the visual resources field and not the bibliographical field was discussed. Also, the wide variety of people and goals represented had presented a problem. Art-historical visual collections are different from other visual collections and need different treatment. For instance, those people indexing photo archives documenting real events need subject terms distinct from those indexing projects cataloguing works of art. Betz's List of subject headings for the Library of Congress Prints and Photographs Division, although intended for visual material, is too close to the bibliographical approach. A new type of terminology is needed, since on the whole visual art has been recorded in terms that are bibliographically rather than arthistorically orientated. Consequently, in the cataloguing of a work of art, the emphasis has commonly been upon cataloguing, indexing and making available the carrier of iconographical information rather than the iconography itself i.e. it is the slide and not the work of art, which has been indexed.