Winner of the 1997 VRA Nancy DeLaurier Writing Award, in the student paper category.
Published: Angeles, M. (1997, Fall). Analysis and Description of an International Visual Resources Index: A proposal for building an image indexing database. VRA Bulletin, 24 (3), 37-59.
Table of contents
5. Documentary units (Units of analysis)
8. Exhaustivity of indexing (analysis and representation)
9. Specificity of indexing terms
10. Displayed versus non-displayed indexes
12. Vocabulary tracking and management
13. Surrogation of messages and texts
14. Locators and links to documents
16. Arrangement of displayed indexes
18. Record structure and examples
Preface
This database design proposal is a result of my humble first effort to explore methods for representing art historical pictorial images from a Library and Information Science perspective. The paper was produced as an exercise in structuring an index to represent knowledge descriptions, the messages of documents, corresponding with a knowledge representation course taken in the Master of Library Studies program at Rutgers University -- School of Communication, Information, and Library Studies. Before presenting the design and analysis of the database, I believe it will help to discuss some of the issues which make the intellectual work of representing image meaning and content a challenging and unique process.
In my pursuit of literature discussing indexing systems to facilitate discovery of visual resources, it became apparent that deciphering the so called "text" of an image, and representing that text in a surrogate made up of descriptive data, is a process fraught with problems and vagaries. Among the questions pertinent that came immediately to mind were, "How will users search for images?", "What level of description and classification is necessary?", and the big one, "Are pictorial images even truly describable?". These issues will be explored throughout the paper.
I have been introduced to the practice of indexing through James D. Anderson's course "Knowledge Representation and Indexing for Information Retrieval", and his accompanying text, "The Design of Indexes for Textual Databases" (forthcoming). Professor Anderson, has instilled the virtues of sound and logical design of database indexes for the discovery of documents, in students of this course, such as myself. He has also encouraged us to explore new methods for designing databases. It is primarily through him, my fellow students, and the literature of this course that I have derived any knowledge in this corner of Library and Information Studies.
The sections which follow are structured according to Anderson's model in the forthcoming text (patterned after NISO Z39.4-199X) defining the practices employed in the design of this database.
Introduction
This document defines the structure of an online, electronic database whose purpose is to represent the messages of art works through pictorial images. The database will particularly represent images of works in the visual arts (often referred to as visual resources). Let us be clear of the distinction that we define the image as the mechanism which serves to picture the object, or event as it exists or existed in the real or imagined world. As such, we are essentially representing the original work in this index through the use of the image.
Indexing can and often does fail searchers. This statement particularly applies to pictorial image indexing, which is made more problematic by the fact that the study of how humans perceive pictures, often referred to as "Visual Literacy" and "Image Communication", is a relatively new area of study that crosses the disciplinary lines of the humanities and social sciences.
Pictorial images are documents which can in many ways defy description and classification. Bellour eloquently sums up the elusiveness of pictorial images in a comment regarding cinematic film. Images, he says, evade us "... by being literally and figuratively unquotable, everlastingly slipping through in the instance of being identified, seized for scrutiny" (O'Connor "Access to Moving Image Documents..." 210).
Representation of pictorial images is fraught with many difficulties. Elizabeth Betz identifies some of the challenges of image representation in the Library of Congress manual "Graphic Materials".
"Graphic materials ... have little or no text to transcribe. The major reason for documenting graphics is to provide the [searcher or] researcher with as complete an identification of the material as possible. This is done by translating the visual information into a verbal description of the material's physical nature and image content. Authenticating the material and making attributions of responsibility are also activities in documenting graphics. Information must be extracted, interpreted, and extrapolated from the visual content and context of the material, as well as from secondary sources" (Betz 4).
Karen Markey states that art historical researchers generally ask five common questions of an art work: "(1) when was it executed (date)? (2) Where has it been since its execution (provenance)? (3) Who created it (responsibility)? (4) How was it created (materials, tools, techniques)? (5) Why was it created (function)?" (155).
Researchers of art history will likely search for some of these elements in an art work which are identified through its "document description", a description of empirical data (title, date, size, artist or creator) about the work. Barnett notes, however, that this information is not as easily obtained as the document description of books. While the bibliographic elements which document a book are generally intrinsic to the book itself (such as the author, title, publication date), the information to document the art work are largely extrinsic to it. Items such as title, creator, stylistic period, etc. are often the scholarly opinions and attributions of art historians.
Researchers may also search by "knowledge description" -- subject content or the message represented by the image (Barnett 200, 201). These aspects will be represented as part of a subject matter description. It is the art historian's objective after interpreting the work, "...to make the historical argument clearly, to document that argument fully, and to give to the audience all that is necessary to make a fair judgment on the matter represented" (Brilliant 129). Thus, the index should additionally present all of this art historical interpretation with citations to the sources which document these interpretations.
Representing the message of pictorial documents with verbal and linguistic descriptions, and grouping descriptions into categories is the primary method we will use to overcome the elusiveness of image description. As stated above, this process will prove to be more fuzzy than the process of representing textual items. Brian O'Connor might add to the problem that "there is no saying just how many words are required to describe any individual picture" ("Pictures, Aboutness..."). For this reason, we will approach the image from as many perspectives as we believe a patron may attempt to access the image. It is hoped that an image will be identifiable and therefore retrievable through at least one of the categories describing the image through various verbal statements made in the document description and knowledge description.
Connections to art works can usually be made after an art historian has done this work of describing the image. However, while classification facilitates access to experienced searchers steeped in the appropriate vocabulary, it may pose a problem to others. The searcher's expression of an information need often represents what the s/he does not know. How, then, is the searcher to retrieve messages in a document, if s/he is not in the same state of knowledge as the indexer. Will the words the searcher uses to express his or her information need match the words that the person representing the document, the indexer, will use? This problem may in part be solved by vocabulary control and management. Implementation of sophisticated iterative and interactive Information Retrieval mechanisms which address such problems may also be imminent, but as yet exist only in systems research and development (Belkin, Oddy, Brooks 61).
So how can we, as indexers, communicate the subject of images in a practical, systematic way? Where will we find the decryption key to the elusive image? Aside from sign language and perhaps symbolic/conceptual written languages such as Chinese, there exists no consistent language for describing pictures. Human indexers, and consequently, index searchers describe the subject of an image by stringing together descriptive human language terms into a surrogate text which serves to represent the image. This is the paradigm for image indexing and searching today.
It is believed that we as yet have no direct knowledge concerning Visual Literacy and Image Communication -- how humans perceive pictorial images (Jörgenson par. 4). I believe our inability to understand how we verbalize what we see, lies at the center of our problem in representing images for retrieval. Because of the inventions of photography, moving pictures, television and now multimedia, humans are beginning to communicate as much with pictorial language as with verbal and written language. Indeed some believe that the time will come, if it has not already arrived, when images will surpass verbal and written language as the dominant form of communication. But let us not think that understanding what images mean necessarily implies that we will have power over images (Mitchell 6, 14). What one person sees in an image can and often does vary widely with what another may see; and who is to say that what one person sees is right, and what another sees is not?
Traditional art history and scientific research of image communication has been isolated and limited in scope. Art history, for example, has historically focused on iconology, iconography, history, and sociological issues, but has avoided "Picture Theory". Psychology, has most commonly centered on issues of visual perception. Visual Literacy, however, demands interdisciplinary efforts and perhaps the defining of a new transdisciplinary field to produce usable theories that deal with the meaning of images.
Rudolf Arnheim, refers to the research of psychologist B.F. Skinner, because it focuses on individual ways of looking and interpreting images, which contradicts the practice of eliminating anomalous, immeasurable aspects as "noise", in art historical and scientific research (Arnheim 178-179). This inductive research paradigm goes against conventional classificatory methods of looking at images. Perhaps radical new ways of looking at images will bring theories useful to disciplines outside of those traditionally associated with interpreting the meaning of images. In my humble opinion, the study of Visual Literacy in humans today is as important as the study of the written languages in the early history of humankind, which communicated concepts pictorially (with hieroglyphic scrawlings, clay tablet cuneiform, and symbolic characters). Perhaps, in the future, this is where we will find some answers to our image indexing troubles.
Because the phenomenon of varying descriptions of what is seen in an image exists, our index will rely on descriptors based on patron-centered vocabulary, as well as on the art historical descriptors we have described above. Technology is also becoming available to automatically index characteristics of images such as patterns of geometric lines, shapes, and colors. Hopefully this will add to our image vocabulary
1. Subject scope
The database endeavors to provide access to images of art works, to a broad array of professional and non-professional researchers. To attain this objective, primary and secondary subject access to images will be facilitated. Because the database serves professional researchers, iconographical terms derived by art historical professionals are a primary point of access. Additionally, we hope that users inexperienced in iconographical analysis may also benefit from the system. Therefore, the database necessitates access to iconological/pre-iconographic subject description, and subject description provided by our users themselves. In the future we will consider the possibility of utilizing machine indexers of pictures.
Representation of the image begins with the intellectual work of identification. The type of document we are representing, for example, the title statement, and the statements of responsibility for the work will be indicated. This method of transcription of image data allows the database to provide book-like physical and intellectual access to the pictorial image surrogate, analogous to the various depths of penetration employed by print documents (O'Connor "Access to Moving Image Documents" 209).
The categories which describe the subject identification and the subject matter/image content are broken down into facets loosely based on Ranganathan's facet frame for representing documents, and the Getty Information Institute, Art Information Task Force (Getty AITF hereafter) document "Categories for the Description of Works of Art" (CDWA hereafter), which is alternatively interpreted in the Visual Resources Association document "VRA Core Categories" (VRA CORE hereafter). Many of the definitions, therefore, are taken directly from these documents, as they represent recommended standards. The Getty AITF defines subject matter thusly.
The subject matter of a work of art (sometimes referred to as its content) is the narrative, iconic, or non-objective meaning conveyed by an abstract or figurative composition. It is what is depicted in and by a work of art. (section 18)
The categories representing document description are listed below. Below each category we will list some typical entries. Following the categories will be one example based on the Leonardo da Vinci Mural Painting Last Supper, which we will continue to use throughout the paper to illustrate how one complete document record is entered.
Questions about the subject of an art work can be posed to the database in one of the following categories.
Objects types*
definition: The kind of object/work of art being described. The use of a controlled vocabulary is recommended
e.g. altarpiece, drinking vessel, handscroll, photograph, portfolio
Techniques*
definition: Placement of the work of art within a formal classification.
e.g. architecture. graphic arts. painting. sculpture.
Medium/Materials:
definition: A prose description of the technique, media, and support of the work of art.
e.g. oil on canvas. egg-tempera paint with tooled gold-leaf halos on panel
Titles*
definition: The title(s) or name(s) given to a work of art, as well as the date(s) when the title was valid.
e.g. Ceramic fruit bowl. Untitled. Pieta.
Larger entity names
definition: The name of a larger entity which an work is a part or component of.
e.g. formerly Parthenon (used to identify the name of the temple from which the sculptural group "Three Goddesses")
Creators*
definition: Information about an individual, a group of individuals, corporate body, or cultural group that contributed to the creation, production, manufacture, or alteration of the work.
Names of persons -- Individuals, Dates associated with persons
e.g. Wren, Christopher, 1632-1723. Michelangelo Buonarroti, 1475-1564. Rollins, Tim, 1955-
Names of persons -- Groups, Dates associated with groups of persons
e.g. group of individuals: Rollins, Tim and K.O.S. (Group of artists).
Creator(s) identity or Nationalities/Cultures/Race*
definition: Characteristics that identify an individual or group that played a role in the creation of a work of art.
e.g. Wren, Christopher, 1632-1723, active in England. Unknown Venetian
Styles/Periods/Groups/Movements, Geographic locations & Dates associated with Styles...
definition: The term(s) identifying a style, historical period, school, or art movement whose characteristics are represented by the work of art. e.g. Baroque, Italy. Fauve.
The next category involves the description of subject matter or image content. The art historian distinguishes three levels of subject matter when analyzing art works, primary subject matter, secondary subject matter, and iconological interpretation (Markey 156). Primary subject matter refers to "...forms such as configurations of line and color". Secondary subject matter refers to "...identification of themes or concepts manifested in images, stories, and allegories". Iconological Interpretation refers to identification of underlying principles which reveal the basic attitude of a nation, period, religion, class, or philosophical persuasion (Panofsky 5-8). Research access to images, however, is primarily facilitated through descriptors of secondary subject matter, which is "...in accordance with the training and knowledge of users of such collections and supports their scholarly pursuits". Additionally, Markey notes that although researchers consider primary subject matter in their analysis of art works, historically these aspects have not been included for subject access because the cost of inputting them would be prohibitive (157). The inclusion of pre-iconographic description may serve to increase subject access to art works, however, especially for non-professionals. We will try to include pre-iconographic descriptions when necessary, but will attempt to deal more with this issue in 12. Vocabulary control and management.
The subject matter/image content facets are the indexer's statement of what the picture is of and what the picture is about, represented in verbal/linguistic terms. Art historical analysis can be characterized by a detached reporting of elements such as is done in iconology. It can additionally draw from images, a variety of meanings associated with iconography. It may not appear possible to fit this descriptive statement into a brief frame of authorized terms, but this is the goal of the indexer. The stages of subject matter indexing occur, therefore, on several levels.
The first (optional) stage requires a generic statement regarding "ofness", the description of what one sees in the picture, i.e. what it is a picture of. This type of description is associated with objectively describing the pre-iconographic stages of an image, before it is recognizable as art. When necessary we will fit the terms of our descriptive statement into the facet frame for subject matter.
The second stage requires a statement of iconographic identification, i.e. what specific, identifiable subjects are represented by the picture. Iconography, is a science which is concerned with naming persons, places or things associated with mythological, fictional, religious, or historical subject matter (Panofsky). We will describe what is happening iconographically in a verbal/linguistic sentence and will fit the terms of our descriptive statement into the facet frame for subject matter. The terms in this description will be chosen from a controlled vocabulary source, such as the Art & Architecture Thesaurus.
Description for Iconological Identification may not be necessary, due to the flexibility of term linkage afforded the database through vocabulary management. If narrow terms such as "Madonna" are linked to broad terms such as "Women", records containing Madonna can be returned when one searches under Women. This process will take place behind the scenes. (More on the treatment of cross references can be found in 12. Vocabulary tracking and management). We will group both types of description, generic and iconographical into one facet frame, Subject matter -- Iconographic and Iconological Identification.
Subject matter/image content
Subject matter -- Iconographic and Iconological Identification
Persons (individuals)
Persons (groups)
Properties of persons
Things
Properties of things
Actions or processes
Attributes of actions, processes, events
Places
Concepts
e.g.
Abstract (a verbal-linguistic description): The Madonna and Christ Child are seen seated in an enclosed garden. The Madonna is holding the Old Testament opened to Isaiah 53 and is gazing at the Christ Child, Who holds a pear and is looking at the viewer, raising His hand in blessing. Around them there appears a marble slab, and a vase of lilies at their feet. There is a landscape and storm clouds in the distance.
Possible terms are listed below.
Persons(individuals): Madonna, Christ child
Properties of persons: Seated
Things: Marble slab, sarcophagus lid, Old Testament, book of hours, pear, vase of lilies
Properties of things: Old Testament opened to Isaiah 53.
Actions, processes, events: Christ Child holding a pear, looking at viewer, raising right hand in blessing, Madonna gazing at child
Places: Enclosed garden with landscape and storm clouds in distance
Concepts: death, virgin birth, original sin, salvation
("Definitions" CDWA par 19)
The third stage is a statement interpreting the iconographic message in an abstract expressing the meaning of the terms used in stage two. It is characterized by an explanatory tone which helps clarify the meaning of the iconographic elements identified. The description will be placed in the subject matter abstract. Because of the number of themes that may occur within one image this should be a free-text field, in which the function or role of distinct terms is not identified, as the above fields are. The context of the terms will remain intact because of their placement in relation to other terms. The alternative would be to divide each interpretive statement into facets echoing the structure of the "Subject matter -- Identification" field.
Subject matter/image content -- Interpretation
Abstract
definition: the meanings or themes represented by the subject matter or iconography of a work of art.
Citations:
definition: record references to bibliographic sources, unpublished documents, or personal opinions that provided the basis for the information recorded above.
The following example illustrates how the record using Leonardo da Vinci, Last Supper is entered. Each facet is placed in a distinct field so that the context of the term is preserved (with the exception of terms in the abstract field). Notice for example that there is a date field for Titles (tit-dat/). The categories above are abbreviated by the first three letters of each word.
Obj/ Mural painting
Tec/ Painting
Med/ Oil and tempera on plaster
Tit/ Last Supper
Tit-Dat/ ca. 1495-1498
Cre/ Leonardo da Vinci
Cre-Dat/ 1452-1519
Cre-Nat/ Italian
Sty/ Renaissance
Sty-Geo/ Italy.
Sty-Dat/ 1400-1500
Sub-Per-1/ Jesus Christ
Sub-Per-2/ Apostles
Sub-Act/ Sitting, Last Supper
Sub-Att/ At a table, engaged in conversation
Sub-Concepts/ Betrayal
Sub-Abs/ Thirteen men Christ and the twelve Apostles are seen at the Last Supper seated around a table engaged in conversation about to eat a meal. They are seated at long table set parallel to the picture plane in a simple spacious room. The highly dramatic action of the painting is made still more emphatic by the placement of the group in the austerely quieted setting. Christ with outstretched hands has just said One of you will betray me. A wave of intense excitement passes through the group as each disciple asks himself and in some cases his neighbor Is it I? His face in shadow Judas clutches a money bag in his right hand and reaches his left forward to fulfill the Master's declaration Behold the hand of him that betrayeth me is with me on the table.
Cit/ Gardner, Helen. Gardner's Art through the Ages 9th ed. vol. II. Ed. Horst de la Croix, Richard G. Tansey, Diane Kirkpatrick. San Diego: Harcourt, Brace, Jovanovich, 1991. 637.
(based on interpretation by Panofsky, Markey 156)
To further facilitate access to images by subject content, we will allow liberal patron-generated verbal/linguistic terms to represent the subject of images in the vocabulary of patrons. This field must be repeatable, as it is intended that entries in this field grow organically with the patrons. These terms can be characterized by idiosyncratic observations directly tied to patron reaction to images, that go beyond the topical descriptions of subject matter. They go to a higher or lower level of specificity beyond topicality at the level of the document (O'Connor Explorations in Indexing... 153). They are likely to include adjectival and functional descriptions of, for example, emotions or a suggested dialogue within the moment pictured.
Subject matter/image content -- Patron descriptors/captions
definition: patron contributed adjectival terms, functional terms
e.g.(1)
confusion
disbelief
Optional fields may be added to the database for future machine indexing which supply iconological descriptors indicating forms such as line, geometric shape, and color by performing pattern recognition on individual images. An example of technology which may be used to machine-index image content are IBM, Query by Image Content (QBIC). Markey states that iconological interpretation of this type is possible, but is probably not necessary for an image database (158).
2. Documentary scope
Format/Techniques: The types of art works represented include, but are not restricted to the following formats and techniques: two-dimensional works such as drawings, photographs, prints, and paintings; three-dimensional works such as architecture and other built environments; built or sculpted sculpture; ephemeral works such as performance art; and kinetic works such as cinematic film. The art works which are represented should not be confused with the surrogates which represent art works through various digital media (see 13. Surrogation of messages and texts for more on the visual media used to serve as reproductions of the original art works.)
Medium/Materials: Materials of art works are any materials used to transform the intrinsic qualities of the material used, to the extent that the material subsequently represents a work or object of art. Examples of these materials include acrylic, ink, digital media, fur, marble, oil paint, wood, video, etc.
Periodicity: Works of art may generally be produced at one point in time, but art works are also produced periodically or serially. Both types will be represented in the database.
Audience or level: Research scholars in the field of art history, at professional, undergraduate and graduate school levels are expected to use the database.
Nationality: Documents may be drawn from any geographic location or nation.
Time: No constraints are placed to exclude works based on date of creation. Documents may be represented from any time period.
Specific documents: The database does not exclude works other than those that do not serve art historical research purposes, and those which do not fall within the realm of the visual arts. Documents will likely be drawn from the art and image collections of private and public art galleries, institutions, and museums.
Qualitative criteria: The qualitative criteria for inclusion of art works in this database will be based on the importance of the art work to the field, as deemed by art historical researchers and selected major museums, galleries, and institutions.
3. Domain
The database points to art works from a wide range of sources. It is intended to be an international repository of art work images and information. This information will be selected by monitoring the collections of museums and institutions internationally, and it is hoped that they will see the benefit of contributing to this database.
Indexing will be based on primary sources and secondary sources. Museums most often index art works that are part of their collection, so these will usually be based on primary sources. However, visual resources collections in academic institutions, and indeed also in museums, often necessitate indexing of works which are not accessible for physical inspection. This should not present an obstacle to indexers, however, because citations to sources of information will usually accompany the indexing data. Researchers can verify that the authority of the record, by checking the citation.
In theory, this database will include art works from all of the world's major galleries, museums, art and cultural institutions. By basing the intellectual structure and design of the database on developing standards (CDWA, VRA Core) it is likely that museums will be able to be share art images and information through the database. If initiatives such as the Consortium for the Computer Interchange of Museum Information (CIMI) are able to establish standards for a museum computer network utilizing communications protocol standards (such as Z39.50), this goal will be possible. However, because of the obstacle and unresolved problem of Intellectual Property, we are not sure if reproductions of all of the images representing art works can be surrogated in this database. Work is being done toward this effort, however, by consortia such as The Museum Educational Site Licensing Project (MESL) (Trant).
4. Display media
The database will be available in electronic form, accessible online through a computer network connection. Producing an online, electronic database allows this repository to be accessible internationally in its updated state in real time, eliminating the obsolescence associated with print and CD-ROM indexes, in that they are outdated once they are produced. This will also mean that database information can be highly manipulable, and there will be many possible ways for data to be displayed. Electronic searching of all document and knowledge facets will be accommodated.
The database will present information in accordance with English language usage, employing the standard Roman alphabet, punctuation symbols, and Arabic and Roman numerals (in accordance with the ISO standard). However, empirical data such as document title will be recorded in the form that is given by the artist, usually in the language of the art work's origin. Whenever non-Roman alphabetic symbols are used, the language of the alphabetic symbols being employed will be indicated (in the Notes field), and the rules for usage and arrangement will follow the standards of that language.
5. Documentary units
Documents will be indexed at the level of the entire art work. Art works refer to the specific paintings, drawings, sculptures, performance art works, digital art works, video, and architectural buildings and environments, that are considered under one title, representing one entire work. Works that are elements of a series will be considered as one unique document at the level of the individual art work or element, and not at the series (which will be referenced in the document description, however, under the field "Larger entity").
6. Indexable matter (Analysis base)
Indexing at the level of the art work will produce the intellectual information (document and knowledge description) of the art work. In presenting each art work as an image surrogate, separate records are required, indexing each image at the level of the image. Handling images in this manner, will allow us to keep a unified document and knowledge description of the work (the documentary unit) in one place, and image surrogates and descriptions of particular image views (e.g. view from below, interior) in a separate location. We also plan to index at the level of the digital bitmap and pixel in the future, using image pattern recognition software. We consider this to be a type of full-text, or full-image indexing.
7. Analysis methods
Most descriptions will be supplied by professional human indexers who are art historical researchers. Additional descriptions will be user generated, and are intended to grow organically with patrons. Finally, some descriptions of images may be supplied by automatic indexers in the future. Therefore, the types of requests that may elicit a response will be of a multifarious nature.
The methods of analysis we are describing is characterized by three distinct approaches. The human indexing methods of the art historians will be thorough, and as such, time consuming and costly. This aspect is essential, however, and is not considered a detrimental cost in terms of the benefit it provides to our intended searchers. We have stated in this introduction that we believe the analysis methods of images to be inadequate to serve all people. However, the analysis methods of art historians is without a doubt, we believe, the most exhaustive and studied approach possible.
To compensate for the problem of anomalous states of knowledge (Belkin) we have integrated user-centered vocabulary in our database. It is hoped that the practice of allowing patrons to enter words which they associate with an art work, or with a particular image of an art work will serve to increase recall. It may be argued that a great deal of precision may be lost using this field, if patrons attach nonsensical words or misplaced words to an image. However, this field may arguably also increase the precision of the database if terms that could not possibly be attached to an image because of the nature of iconographic analysis. The effect of patron-centered indexing depends upon the terms that are used. It is believed that terms will begin to cluster around those documents which are retrieved commonly. Certain of these terms may begin to penetrate the meaning of the message of the image more deeply, and therefore increase precision in the database. All of this will remain to be seen. Indeed, the use of this field may be seen as a bother, in which case, nothing will be lost. This further exemplifies the nature of this field as an added value to the database.
The automatic indexing of images may also help to increase recall of documents. New image vocabulary terms may arise that describe the image in alternative fashion. This field is also meant to be an added value. We have included both of the latter non-art historian/human indexing methods to increase the number of approaches to art works.
8. Exhaustivity of indexing
Because there are many approaches to image description, the detail of indexing may seem to demand a fairly high level of exhaustivity. When compared to text indexes, image indexes cannot be cleanly classified as exhaustive or inexhaustive, because "detail of image information" is an elusive concept. Clearly, one could describe an image in detail, ad infinitum, or could sum it up in a few words. In some of the indexes we will call for high levels of exhaustivity and, and moderate levels in others.
Document description can be exhaustive if we include all of the details concerned with an art work's history. For example, we can include all of the creators of an artwork or we can follow the "Rule of three" method employed in Library descriptive cataloging. The advantage of exhaustive indexing is that the database will allow searchers to excavate the fine details which advanced researchers so often will want to find. The database will attempt, therefore, to index documentary data at a moderate to high level of exhaustivity.
We believe, because we intend to serve advanced researchers, that users of the database will usually want the most relevant documents returned in a search (low to medium recall characterized by a high-threshold of importance) of knowledge descriptions (subject matter/image content). But, there may be many possible terms to describe knowledge descriptions (moderate to high level exhaustivity). We will discuss the specificity of term application in the next section.
9. Specificity
Our indexing practice will be guided by a method of "practical specificity", using the narrowest descriptors, which most closely fit the central message of each document at the level of the work of art, that fits without getting so specific that no-one will find the headings (Anderson sect. 2.9). The database additionally provides access from a variety of perspectives to ensure that access to documents is not hampered by the specificity of terms. We hope that vocabulary control (see 12. Vocabulary tracking and management) will allow users to increase recall through cross references to broader, narrower or related terms.
We will illustrate specificity by returning to the Leonardo example. In the image of The Last Supper, thirteen men are seen sitting around a table, engaged in conversation around a man at the center of the image who faces the viewer. They are about to have a meal. This description uses broad terms as a description of the image's subject content, but this event will not likely be searched as a broad topic using terms such as "thirteen men" or "conversation". In this case, we are dealing with a narrow event, so we can use narrower terms, describing it as, "Jesus Christ seated with his 12 Apostles at the Last Supper". Or we can be more specific by mentioning that it is the scene that is commonly known by subject, as "The Last Supper". And there may be more that we can say to identify other specific aspects of the image. We can discuss the detail of Judas holding a sack of coins. But is this element, while it captures the subject of the moment pictured most symbolically, the central message of this image? It is the act of the "Last Supper", which most tightly fits the meaning of this image. There is a lot of meaning in this image, however, so we will not exclude the other important activities, because they will be highly valued aspects which researchers will most likely search for at some point. Aspects such as the description of Judas and the identification of what the gesture of grasping the sack of money means will be included in the Subject matter/image content -- interpretation abstract (see 1. Subject scope). A researcher demanding high recall can search by the terms in this expansive field.
10. Display of index
The database is intended to serve experienced researchers, who may know exactly what they are looking for, and to serve researchers who may only have a general idea of what they are looking for. Therefore, we will give the searcher the ability to search using a non-displayed index, or to visually search a displayed index which orders the subject matter terms of a category in an alphabetical list.
Displayed indexes will allow the searcher to visually browse or scan through a list of headings. This type of searching is common to print indexes. An index will usually list headings alphabetically or in classified order, or in a combination of both. Selecting a category will create an alphabetical list of that category "on the fly", for user browsing. The user of the index will search for a heading and browse the attached subheadings which modify the meaning of the lead term. The user of this database will be able to go to a displayed index for subject matter, for example, following the same paradigm, where s/he can search for a heading, and scan the terms below the lead term to find the document that is most relevant to their search.
If the user chooses to go directly to documents which are indexed by a certain term or terms, they can enter a search statement into a query without seeing the index. This search statement will pose a question to the database. The documents which answer the question can be returned in a variety of ways, usually displaying a list.
Offering two approaches to what users see on the database screen when they are searching ensures that varying types of searchers, and varying search styles are accommodated.
11. Syntax for representation
This section discusses the rules guiding the arrangement of the terms (the "syntax") we have attached to the image in section 1 Subject content. The independent terms offer scanty insight into the message of the image. The combination of terms, however, makes a statement of the unique message of the image.
In the displayed index, the combination of terms will follow "faceted citation order syntax", similar to the type of indexing done in the Modern Language Association Bibliography. This method takes the terms which indexers have placed in the subject matter facet frame above, and arranges them into an ordered string. It is characterized by a high level of precision, which ensures that the context of one term in relation to the other terms in the string, is preserved. Additionally, it allows users to keep recall low, by being able to eliminate documents that are irrelevant. Users will be able to increase recall by massaging the search based on broader or related terms found by virtue of cross references.
The display of headings in the subject index will be generated on-the-fly. Each of the terms from the facets take the lead position. The database takes each of the facets and arranges them into the display below the heading (the lead term) based on facet order. The entire string of Subject matter/Image content terms are repeated in facet order at the end of the document description entry. The lead term is repeated in UPPERCASE in the display. Following the display terms, there will be a hypertext link, analogous to a locator, that will cause the surrogate of the entry to appear on the screen. The order of each entry will assume the following pattern, the first line in called the Lead term, the section following is called the display
Lead term from subject matter/image content
Style, geographic location-dates. Creator, dates, nationality.
Work title. Technique. Object. Medium.
Subject matter/image content terms.
[surrogate link]
("extended citation" display format)
We will illustrate the advantage of faceted order syntax, once again using Leonardo's Last Supper.
Apostles
Renaissance, Italy. 1400-1500. Leonardo da Vinci, 1452-1519.
"Last Supper", ca. 1495-1498. Painting. Mural Painting. Oil and tempera on plaster.
Treatment of Jesus Christ. APOSTLES. Last Supper. Betrayal.
[surrogate link]
Betrayal
Renaissance, Italy. 1400-1500. Leonardo da Vinci, 1452-1519.
"Last Supper", ca. 1495-1498. Painting. Mural Painting. Oil and tempera on plaster.
Treatment of Jesus Christ. Apostles. Last Supper. BETRAYAL.
[surrogate link]
Jesus Christ
Renaissance, Italy. 1400-1500. Leonardo da Vinci, 1452-1519.
"Last Supper", ca. 1495-1498. Painting. Mural Painting. Oil and tempera on plaster.
Treatment of JESUS CHRIST. Apostles. Last Supper. Betrayal. [link]
Last Supper, ca. 1495-1498. [subject]
Renaissance, Italy. 1400-1500. Leonardo da Vinci, 1452-1519.
"LAST SUPPER", CA. 1495-1498. Painting. Mural Painting. Oil and tempera on plaster.
Treatment of Jesus Christ. Apostles. Last Supper. Betrayal.
[surrogate link]
Last Supper. [title]
Renaissance, Italy. 1400-1500. LEONARDO DA VINCI, 1452-1519.
"Last Supper", ca. 1495-1498. Painting. Mural Painting. Oil and tempera on plaster.
Treatment of Jesus Christ. Apostles. LAST SUPPER. Betrayal.
[surrogate link]
Leonardo da Vinci, 1452-1519.
Renaissance, Italy. 1400-1500. LEONARDO DA VINCI, 1452-1519.
"Last Supper", ca. 1495-1498. Painting. Mural Painting. Oil and tempera on plaster.
Treatment of Jesus Christ. Apostles. Last Supper. Betrayal.
[surrogate link]
Mural Painting.
Renaissance, Italy. 1400-1500. Leonardo da Vinci, 1452-1519.
"Last Supper", ca. 1495-1498. Painting. MURAL PAINTING. Oil and tempera on plaster.
Treatment of Jesus Christ. Apostles. Last Supper. Betrayal.
[surrogate link]
Oil and tempera on plaster.
Renaissance, Italy. 1400-1500. Leonardo da Vinci, 1452-1519.
"Last Supper", ca. 1495-1498. Painting. Mural Painting. OIL AND TEMPERA ON PLASTER.
Treatment of Jesus Christ. Apostles. Last Supper. Betrayal.
[surrogate link]
Painting.
Renaissance, Italy. 1400-1500. Leonardo da Vinci, 1452-1519.
"Last Supper", ca. 1495-1498. PAINTING. Mural Painting. Oil and tempera on plaster.
Treatment of Jesus Christ. Apostles. Last Supper. Betrayal.
[surrogate link]
Renaissance, Italy. 1400-1500.
RENAISSANCE, ITALY. 1400-1500. Leonardo da Vinci, 1452-1519. "Last Supper", ca. 1495-1498. Painting. Mural Painting. Oil and tempera on plaster.
Treatment of Jesus Christ. Apostles. Last Supper. Betrayal.
[surrogate link]
12. Vocabulary tracking and management
Many of the fields in which terms which are chosen in the description of the art work are based on controlled vocabulary sources. The following list indicates which sources were consulted when choosing terms in the categories specified.
| Category | Controlled vocabulary source |
| Objects types | AAT (objects facet) |
| Techniques | AAT (processes and techniques hierarchy) |
| Medium/Materials | AAT (materials hierarchy) |
| Titles | No controlled vocabulary source, the preferred name used by the repository is entered |
| Names of persons -- Individuals | ULAN |
| Names of persons -- Groups | ULAN |
| All geographic locations | TGN |
| All dates | ISO 8601 Standard |
AAT, Art & Architecture Thesaurus. TGN, Thesaurus for Geographic Names ULAN, Union List of Artist Names. (VRA CORE)
The use of a thesaurus will 1) bring together equivalent terms expressed in a variety of ways, 2) link terms which are closely related, and 3) link terms to broader and narrower terms within the subject of the term being browsed. In addition, vocabulary control will specify the differences between homographs, terms identically spelled, which represent different ideas, by attaching a modifying term in parenthesis (e.g. Mercury (god), Mercury (planet)). As we have discussed in 11. Syntax for representation, the displayed index incorporates cross references from the thesaurus directly into the display. When browsing a displayed list, users will be able to visually see that unauthorized terms are not used, and will be led to the equivalent, authorized term through a hypertext link. Users will also be able to browse entries related to the term retrieved.
The non-displayed index will handle vocabulary control quite differently. In the search screen, a user will be able to query the database by specific fields or by keywords. Before s/he enters a search statement, however, s/he will also be able to indicate whether or not s/he wishes the terms in the statement to be compared against a controlled vocabulary. Additionally, the user can choose among two methods of vocabulary control checking. In "default mode", the thesaurus will return all variations of a term (e.g. searching under the term "horse" will return "horses" as well). In "browse mode", any authorized terms which include the requested word will be returned. The user then has the option to choose from an alphabetically ordered list, the term which applies the their search. The results of the search will be retrieved in ranked order. Users who chose to search without vocabulary control will retrieve exactly the term which is entered.
The use of a thesaurus serves to decrease recall of irrelevant documents. The linkage of equivalent terms ensures that users retrieve all documents relating to one idea (i.e. precision increases and recall therefore decreases). The linkage of related terms ensures that searchers can retrieve documents which may fall within their search need, that are represented by terms they may not have thought of in doing their search. Although these processes are iterative and require extra steps, we believe the integration of the thesaurus into the system will increase the precision of searches and reduce the recall of irrelevant documents.
13. Surrogation of messages and texts
The document surrogates in the database will consist of a lengthy document description, the full knowledge description (see 1. Subject scope), and a digital image surrogate. In this section we are once again turning to the Getty AITF CDWA and the VRA CORE as the guidelines by which we are deriving our bibliographic statements about art works. We will primarily be using the CORE elements of the CDWA. The CDWA is consulted because it "... is intended to be a statement of research needs of scholars, prepared by researchers themselves and by those who provide information to researchers" (par. 55).
It is likely that most researchers will search for images directly by the categories in the surrogate. The surrogate frame below indicates the types of elements that can be put into the surrogate. The elements are divided into four categories: 1) Object description, 2) Creator description, 3) Knowledge description, and 4) Image surrogate description. The art work will be represented by various digital media including still or kinetic photographic (and possibly audio) media. These surrogates will be held in a container field that holds the digital surrogate. Elements which were defined above are indicated by asterisks.
Object description
Objects types*
Techniques*
Medium/Materials*
Titles*
Larger entity names*
Repository name
Repository place
(USE TGN)
Repository number
definition: Identification of the repository that currently houses the work of art, its geographical location, repository number.
Site
definition: the geographic place where an site specific work is located or was found or placed. (USE TGN)
Notes
definition: Notes regarding any of the object attribution information including citations to sources of information.
Creator description
Names of persons -- Individuals*
Names of persons -- Groups*
Dates associated with persons*
Creator(s) identity or Nationalities/Cultures/Race*
Styles/Periods/Groups/Movements*
Geographic locations & Dates associated with Styles...*
Knowledge description
Subject matter/image content Iconographic and Iconological
Identification terms*
Subject matter/image content -- Interpretation Abstract*
Citations*
Digital image container (thumbnail)
definition: Container field which holds a thumbnail digital image of the art work being represented.
Digital image container (full image)
definition: Container field which holds a high resolution digital image of the art work being represented.
Digital image description
View description
definition: Terms or phrases which describe the specific view or perspectival description of how the art work has been captured in the image, e.g. "View from below", "Interior view"
Image types
definition: Terms which describe the digital format of the image. e.g. "bmp", "jpeg", "tiff", "quicktime movie"
Image owners
definition: The identification of the public or private repository or individual who owns the image, including the geographic location.
Owner numbers
definition: The unique identification number used by the above repository or individual.
User defined keywords
definition: Keywords supplied by users at the level of the image.
Machine defined keywords (for future consideration)
definition: Keywords supplied by automatic indexers at the level of the image.
14. Locators and links to documents
Locators occur in the database in three forms: 1) The hyper text link, 2) The digital image, and 3) The full document surrogate.
The hyper text link is a locator that is found throughout the index. For example, in 11. Syntax for representation, we demonstrated the display of the subject index. Following each display was something that read [surrogate link]. This code was meant to indicate that a button could be selected to link to the full document surrogate. We will use hyper text links in various areas of the index, such as in the cross-references. We will indicate that a button or a segment of text is associated with a hyper text link by using blue colored text and buttons -- the color blue is often associated with hyper text links on the World Wide Web.
The digital image is a surrogate that works on two levels. The thumbnail image points to a higher resolution of the image. Both images serve as locators that indicate a real work of art.
Finally the entire document surrogate that is represented by the text and images discussed in 13. Surrogation of messages and texts, serves as a locator that indicates a real work of art.
15. Surrogate displays
While searches on the database can result in varying views of lists, all searches eventually lead to the full surrogate of the art work in the unified display. The unified display lists all of the elements defined above. The combination of descriptions serves to identify the work of art as a unique item. From this display, the user has the option to look at images. The unified display will show an thumbnail image or a list of a number of thumbnail images. If there are more than one image, all of these images will appear in the unified display. When the user clicks on the image, a display of the full resolution image will appear, with its accompanying image surrogate descriptions.
Throughout the researcher's use of the database, s/he will be able to specify how they would like data to displayed. Ordinarily, a scrollable list will be generated on the fly, with links (locators) from these lists leading to each item's location in the unified display. If the user decides to define the display of results, the items located will be listed as requested. For example a user can choose among a set of pre-formatted views from a pop-up menu, named "brief citation", "standard citation" (the default), or "extended citation". The brief citation will list lead terms, and knowledge description terms (subject matter/image content strings). The default view, the standard citation, will list everything in the brief citation, plus document descriptions such as creator and title information. The final, extended citation will list everything in the standard citation plus style, geographic location-dates, technique, and medium. Users can alternatively select the categories they would like displayed in the search results list. (see illustration in A-3 Vocabulary control and display options)
16. Arrangement of displayed indexes
Users of the database have the option to browse displayed indexes by any of the "document description" or "Subject matter" categories displayed (see Appendix A-2, for a view of the opening screen display). The standard method of browsing will allow the searcher to select a category to visually scan. The beginning of the standard format of the displayed index will appear in alphabetical order. A "Jump" window will appear to the side of the display, where the searcher can begin entering a term (that fits into the chosen category scope) into a data entry box. The index will jump to the section of the display where that terms occurs. The user can specify an alternative display instead of using the default standard display format. (See Appendix A-3, Vocabulary control and Display options for selection of alternative displays) Beneath each heading the user will see the display format selected, followed by a button showing the link to the surrogate of the document being represented.
Lead term (from any category)
Creator, dates, nationality.
Work title.
Subject matter/image content terms.
[surrogate link]
("standard citation" display format, showing an entry for one document)
The following example illustrates how the displayed index would appear if the user chose to search for paintings in Italy that have "Jesus Christ" and "Apostles" as subject matter. The user here wanted only to display "Geographic location" and "Technique" categories, so there is little information about what the surrogate links lead to. For this reason, at the very least, the default display above is recommended because it shows the description of the art work in a fuller format.
Jesus Christ. Apostles.
Italy.
Painting.
[surrogate link] [surrogate link] [surrogate link] ...
(user specified display)
In the example above, the user can see the number of documents that reference the returned heading by looking at the number of surrogate links shown. Contrast this to the search of a non-displayed index, where the query will return a number that will tell the user how many documents fit the search, and subsequently will show a list of those documents in the format requested.
The term headings are ordered alphabetically by the lead term. For the most part, the database will follow the faceted-order syntax mentioned above when sorting documents. This ensures that a sense of consistency exists when scanning the displayed index. If the user specifies that only certain categories will be viewed in the display, the computer will re-order the index entries based on the lead term, and will do sub-ordering of terms by the terms in the second position, and then by every term in the subsequent positions. The computer will be programmed to follow the rules governed by the ALA filing rules.
17. Search interface
When a user launches the database, s/he will be greeted by an opening screen that gives some brief information about the contents of the database and lists some options. The user may choose to 1) do a quick keyword search that looks through all terms in the database, 2) do a search of terms occurring in one of the document or object description categories or by one of the knowledge/subject description categories, 3) go to a screen that allows advanced term searching of multiple fields, or 4) browse a list of topics by category in the displayed index. (see illustration in Appendix A-1, Opening screen display, and A-2 Displayed index browsing)
In the search of the displayed index (option 4), cross references to broader, narrower, and related terms will be placed directly below headings (lead terms) and above the display. Each cross reference will be indicated in boldface, and in a color indicating that it is a hypertext link. When pressed, a new list will appear where the cross referenced term occurs in alphabetical order within the index.
e.g. unauthorized term
Lead term. ============================ for information on this topic, see: Authorized term link ============================ e.g. authorized term with cross references to related terms Lead term. ============================ you may view broader terms: Term link you may view narrower terms: Term link you may view related terms: Term link ============================ Display terms. [surrogate link] Display terms. [surrogate link]
In the non-displayed index, users will be able to search for terms that may either be extracted from the document or knowledge description, or in any of the additional fields which will be described in the surrogate, including current location of work, image format, etc. The user may also search by a keyword, which can occur in any field. As such, the system will search for the term or terms the user specifies in their query and will generate a customized list of documents which contain the terms, based on the type of query posted. All of the document entries will be generated on-the-fly after the user enters a search statement (postcoordinate syntax). These documents will be returned by ranked order of importance (based on term weights and frequency of document use). The user will alternatively have the ability to do "exact match" or Boolean searches using logical operators AND, OR, NOT. (see illustration in Appendix A-1 Opening screen display and A-4-6 Advanced multiple field searching)
The non-displayed index will also allow the user to use vocabulary management techniques in an intermediary step after the search statement is entered. Discussion of these methods are discussed in 12. Vocabulary tracking and management. (see illustration in A-3 Vocabulary control and display options)
18. Record structure
The complete record structure that the database uses is illustrated below, using the Leonardo Last Supper, once again.
Obj/ Mural painting
Tec/ Painting
Med/ Oil and tempera on plaster
Tit/ Last Supper
Tit-Dat/ ca. 1495-1498
Cre/ Leonardo da Vinci
Cre-Dat/ 1452-1519
Cre-Nat/ Italian
Sty/ Renaissance
Sty-Geo/ Italy.
Sty-Dat/ 1400-1500
Sub-Per-Ind/ Jesus Christ
Sub-Per-Gro/ Apostles
Sub-Act/ Sitting, The Last Supper
Sub-Att/ At a table, engaged in conversation
Sub-Con/ Betrayal
Sub-Abs/ Thirteen men, Christ and the twelve Apostles, are seen at the Last Supper, seated around a table, engaged in conversation, about to eat a meal. They are seated at long table set parallel to the picture plane in a simple, spacious room. The highly dramatic action of the painting is made still more emphatic by the placement of the group in the austerely quieted setting. Christ, with outstretched hands, has just said, "One of you will betray me." A wave of intense excitement passes through the group, as each disciple asks himself and, in some cases, his neighbor, Is it I? His face in shadow, Judas clutches a money bag in his right hand and reaches his left forward to fulfill the Master's declaration: "Behold, the hand of him that betrayeth me is with me on the table."
Cit/ Gardner, Helen. Gardner's Art through the Ages 9th ed. vol. II. Ed. Horst de la Croix, Richard G. Tansey, Diane Kirkpatrick. San Diego: Harcourt, Brace, Jovanovich, 1991. 637.
Rep-Nam/ [repository name listed here]
Rep-Pla/ [repository place listed here]
Sit/ Santa Maria delle Grazie, Milan, Italy
Not/ [notes listed here]
Img-Thu/ [linking reference to image thumbnails listed here]
Img-Ful/ [linking reference to full image listed here]
Usr/ confusion. betrayal. [user defined keywords listed here]
ID/ [unique primary identification key listed here]
Works cited
1) American Library Association, Filing Committee. ALA filing rules. Chicago: American Library Association. 1980.
2) Anderson, James D. Indexing for Information Retrieval. Forthcoming.
3) Arnheim, Rudolf. "A Plea for Visual Thinking". The Language of Images. Ed. W.J.T. Mitchell. Chicago: University of Chicago Press, 1980. 171-179.
4) Art & Architecture Thesaurus (AAT) 2nd ed. New York: Oxford University Press.
5) Art Information Task Force, Getty Information Institute. "Categories for the Description of Works of Art". Getty Information Institute Home Page (1996): n. pag. Online. Internet. 19 November 1996 Available: http://www.ahip.getty.edu/gii/cdwa/
6) - - -. "Categories for the Description of Works of Art, Definitions". Getty Information Institute Home Page (1996): n. pag. Online. Internet. 19 November 1996 Available: http://www.ahip.getty.edu/gii/cdwa/DEFDOC.HTM
7) Barnett, Patricia J. "An Art Information System: From Integration to Interpretation". Library Trends 37 (2), Fall 1988. 195-205.
8) Betz, Elisabeth W. Graphic Materials: rules for describing original items and historical collections. Washington, D.C.: Library of Congress, 1982.
9) Belkin, N.J., Oddy, R.N., Brooks, H.M. "Ask for Information Retrieval: Part I. Background and theory". Journal of Documentation 38 (1982): 61-71.
10) Brilliant, Richard. "How an Art Historian Connects Art Objects and Information". Library Trends 37 (2), Fall 1988. 120-129.
11) Consortium for the Computer Interchange of Museum Information (CIMI). Online. Internet. Available: http://www.cimi.org/cimi/
12) International Organization for Standardization. ISO 639: Language Codes. Codes for the Representation of Names of Languages/ Codes pour la représentation des noms de langue. 1st ed.[Geneva, Switzerland]: International Organization for Standardization, 1988.
13) Jörgenson, Corinne. "Indexing Images: Testing an image description template". ASIS 1996 Annual Conference Proceedings. October 19-24 (1996). 30 pars. Online. Internet. 18 April 1997. Available. http://www.asis.org/annual96/ElectronicProceedings/jorgensen.html.
14) Markey, Karen. "Access to Iconographical Research Collections". Library Trends 37 (2), Fall 1988. 154-174.
15) Mitchell, W.J.T. Picture Theory. Chicago: University of Chicago Press, 1994.
16) O'Connor, Brian C. "Access to Moving Image Documents: Background concepts and proposals for surrogates for film and video works", Journal of Documentation 41 (1985): 209-220.
17) - - -. "Pictures, Aboutness, and User-generated descriptors". The SIG VIS News 1 (2) (Spring 1996) n. pag. Online. Internet. 19 April 1997. Available http://www.unt.edu/~aag0001/oconnor.html
18) - - -. Explorations in Indexing and Abstracting: Pointing, Virtue, and Power. Englewood, CO: Libraries Unlimited, 1996.
19) Panofsky, Erwin. Studies in the Visual Arts. New York : Harper and Row, n.d.
20) Thesaurus of Geographic Names (TGN). Forthcoming from the Getty Information Institute.
21) Trant, Jennifer. "Exploring New Methods for Administering Intellectual Property: The Museum Educational Site Licensing (MESL) Project". Proc. of 33rd Annual Data Processing Clinic, University of Illinois at Urbana Champaigne, March 25, 1996: Graduate School of Library and Information Science UIUC Publications, 1996 (at press): 45 pars. Online. Internet. Available http://www.io.org/~jtrant/illinois.html
22) Union List of Artist Names (ULAN). James M. Bower, project manager; Murtha Baca, senior editor. New York: G.K. Hall on behalf of the Getty Information Institute, 1994.
23) Visual Resources Association. "VRA Core Categories". VRA Home Page (1997): n. pag. Online. Internet. 6 May 1997 Available: http://www.oberlin.edu/~art/vra/core.html
Thursday, May 29, 1997