Librarian cornered by images or How to index visual resources
Wanda Klenczon, Paweł Rygiel
Wanda Klenczon, librarian at the National Library of Poland (1987- ), Head of Bibliographic Institute (2007- ), chair of Technical Committee Information and Documentation of Polish Standardization Committee. Responsible for the development of bibliographic standards used in the Polish National Bibliography, authority control and subject indexing (National Library of Poland Subject Headings, UDC).
Paweł Rygiel, librarian at the National Library of Poland since 2008. Interested in image indexing and retrieval.
Abstract
Presenting museum, library and archive collections online has become a common practice in recent years. Among the data accessible via the Internet are visual resources such as paintings, drawings, engravings and photographs, which constitute a rich and vital source of information. Image indexing is an extremely difficult task. The crucial question is: how to transform the visual code of an image into written code. This paper presents standards of still images description, guidelines and indexation tools, advantages and disadvantages of their usage, their potential to provide authoritative and comprehensive access points to digitized resources and to support sophisticated search. In order to provide efficient access to the digitalized images presented at the National Digital Library Polona and to promote their re-use, we are looking for a model of images subject indexing.
Introduction
Digital technologies and the online presentation of digital images collections by libraries, museums and archives have enabled wide, absolutely democratic access to resources that were previously available only for researchers. Among the data accessible via the Internet are visual resources such as paintings, drawings, engravings and photographs, which constitute a rich and vital source of information about art, but also about history of people and places, culture and tradition, social life, interests and customs… “While some image collections may be of interest only to researchers in a narrow field […], other image collections are being created that will be of interest to many, including the artist, the scientist, the educator, the historian, the journalist, the student, and the leisure seeker” (Jörgensen, 2003, p. 1).
Availability of both digitalized images and their descriptions has placed them in a new context, while the ways of searching for information have also evolved (Jörgensen, 2003; Jansen, 2008). Moreover, descriptions have been subject to assessment and verification by users who access the freed resources. Collections and institutions may take advantage of this situation as it positively affects the quality of descriptions and helps complete the missing data (see Springer et al., 2008).
Digitalization of images requires a lot of time, money and effort. An extremely difficult task that the cultural heritage institutions have to undertake is to assign high quality semantic metadata to the non-textual materials. Although the automatic recognition of images seems to be a very promising perspective, there are layers of meaning that can be indexed only by using human knowledge, experience and intuition. “Properties of an image such as shape, texture, and color contribute to our understanding of an image, but do not define it. Text-based search techniques remain the most efficient and accurate methods for image retrieval”. (Jörgensen 1999, 302; Neugebauer 2005).
The National Library of Poland as the central library of the state and one of the most important cultural institutions in Poland is responsible for the preservation and promotion of the Polish cultural heritage. The National Digital Library Polona[1] is the repository and the platform for enhancing access to the NL’s digital collections. The number of images (such as prints, drawings, photographs, postcards, posters, bookplates) among the digitized documents constantly increases. In order to provide efficient access to the digitalized images and to promote their re-use, a model of cataloguing and subject indexing should be established. The main goal is to enhance access to both general and specific depicted objects, regardless of the knowledge level of users and their information competency. And last but not least, our intention is to achieve semantic interoperability between the various data and subject schemes used in the Polish libraries and museums communities. In this perspective a few questions should be answered: what vocabulary should be used? What description level would be useful? How could research results be maximized?
Image indexing and retrieval – rules, tools and questions
Subject indexing consists of the following sequence of steps: subject analysis, selection of information and its translation into the language of indexing i.e. transforming the message by changing its language without the loss of information, as far as possible (Chu & O’Brien, 1993; Lancaster, 1998). In the case of iconographic documents it also involves transforming visual code into written code, visual information into textual description, and “the process of translating the content of an image into verbal expressions poses significant challenges to concept-based indexing“(Matusiak, p. 285).
The subject analysis of a language resource can be defined as a process of establishing the subjects of a document. During the process of subject analysis the most suitable subject headings (or descriptors) for the comprehensive description of its content are found. Such analysis is not an easy task. The question is which features of documents should be considered in the subject analysis? (Albrechtsen, 1993). Visual materials may require a completely different approach. Certain features, like title, author, appearance, can be easily identified and used in the description as access points. Furthermore, any image can be differently perceived and interpreted by various groups of users, who may have varying competencies and specific needs.
First of all, in thinking about image indexing, it is necessary to ask if the description of the subject and content of an iconographic document is at all possible. If so, how should it be described? What is the subject of such a document? What is it about? How should works of art be interpreted and what should be included in their description? What aspects of images may be interesting to users?
The problem of analyzing the meaning of a work of art has been discussed by the art historian, Erwin Panofsky, who defined three levels at which a work of art could be described. He suggested that there might be three strata of the meaning of an art work. A work of art can be described on the pre-iconographic level, which consists of general aspects – the primary or natural subject matter; the iconographic level, which encompasses concrete aspects – the secondary or conventional subject matter; and the iconological level, which includes symbolic, abstract aspects – the intrinsic meaning or content (Panofsky, 1962). These three strata of interpretation (pre-iconographic description, iconographic analysis and iconological interpretation) require specific skills from the interpreter, namely practical experience, the knowledge of resources, themes and concepts as well as synthetic intuition (Panofsky, 1962).
Panofsky’s scheme was enriched by Sara Shatford, who also adapted her ideas to image indexing (Shatford, 1986). She renamed Panofsky’s strata as generic, specific and abstract and subdivided each of these areas into four aspects, each of which forms an answer to one of the following questions: Who?, What?, Where?, When?. In this way she created a model of description that encompasses all the aspects of a picture. According to the Panofsky/Shatford model, there are three levels of analysis and description:
-pre-iconographical (generic things, such as woman, flower, building),
-iconographical (specific things, such as Salome, rose, temple), and
- iconological (symbolic or abstract things, such as heaven, love, New Jerusalem).
For example, a flower may be depicted as an element of inanimate nature, simply for its beauty, and it may also appear as a symbol or an element of an iconographical program.
Shatford also introduced the distinction between OF (what a picture consists of) and ABOUT (what a picture is about). The OF/ABOUT paradigm was a critical change that Shatford made to Panofsky’s theories, bringing iconography out of the realm of Renaissance art and into the broader realm of images in general.
The process of indexing is connected with the concepts of ofness and aboutness and isness. Aboutness is the more widely used one of the two terms. It focuses on what is conveyed in a document, what it is about, what its content, subject or theme is. Aboutness depends on the interpretation of the set themes, motifs, actions and events in a work. It is one of the many terms used for expressing certain features of a text or document (features different from the form and the so called description data). Aboutness (and other terms with the same meaning), as a vital element of the organization of knowledge and information retrieval, has been discussed by many researchers (see Hjørland, 2001).
Aboutness is a category that can be used for both language and non-language resources, while the other category – ofness – is a term which only applies to the analysis of pictures. Ofness refers to the elements that a picture consists of. The concepts of aboutness and ofness used in picture description are based on Panofsky’s theory about the three strata of meaning in a work of art. Aboutness and ofness are characteristic of the pre-iconographic and iconographic levels of description. Panofsky’s third stratum – the iconological level of description – refers to the interpretation of the inner meaning or content of a work of art (Panofsky, 1962; Shatford, 1986; see also Zeng, Žumer, & Salaba, 2010).
Isness is another concept which appears in the process of indexing. It tells us what the resource is. Very often this element is close to genre/form terms. The isness of a resource can be physical or intrinsic, e. g. textbook, DVD, poetry, libretto, etc. (Ingwersen, 2002).
Core principles for describing and indexing images have been discussed and agreed upon at the international level by art historians and information professionals (with the predominant role of American experts). The fundamental set of rules and recommendations in this area is known as CCO, Cataloging Cultural Objects. A Guide to Describing Cultural Works and Their Images (Baca et al. 2006). It enumerates the mandatory elements of description, such as:
1. Object Naming,
2. Creator Information,
3. Physical Characteristic,
4. Stylistic, Cultural and Chronological Information,
5. Location and Geography,
6. Subject,
7. Class,
8. Description.
This is a general frame, the basic standard of data content, but the choice of appropriate indexing tool and the decision about indexing policy must be taken by each cataloguing agency respectively according to its collection profile and its users’ needs.
There are two crucial questions to be answered while deciding about images indexing:
- what level of description should be adapted? What is our intention: to express what the image represents (the OF aspect in both a specific and generic sense), or what the image is about (the ABOUT aspect), or to index both of them?
- what indexing tool (tools) could be used successfully? Controlled or uncontrolled vocabulary or both of them? Internationally approved thesaurus or local information language?
While creating an image description, the cataloguer may consider the formal aspects of the object (such as the medium, the technique and/or the material) and other categories that are not directly related to the content of an image. These may include:
- type of document/field of art (photography, engraving, drawing, painting)
- technique/material (collage, oil on canvas, woodcut, ivory carving, photogravure)
- dating (17 century, 1201-1300, the end of 15 century)
- culture, epoch, art movement (Dutch art, Ming Dynasty painting, Art Deco)
However, the indexer’s main task is to specify content features and to create the description of the “content” of a document. The basic elements of content include, among others:
- objects and actions (a girl, a cottage, a boat, a meadow, a bird, sowing)
- type of picture (sketch, self-portrait, still life, veduta, seascape)
- scene or iconographic type (Adam and Eve, Madonna Lactans, Last Supper, Cupid and Psyche, Buddha Amitabha).
Subject analysis and the description of “content” of iconographic documents are sources of numerous problems (Svenonius, 1994; Roberts, 2001). Establishing and defining the subject and meaning of an image may be problematic and identifying the objects is often difficult. Furthermore, the indexing of multifaceted, often allegoric or symbolic images, which are interpreted on many different levels, presumes a certain degree of knowledge of iconography (Schroeder, 1998; Jacobs, 1999). Abstract images, images related to a foreign culture, and images created in different epochs each may contain elements of unknown meaning, indecipherable symbols and gestures, are another source of problems. Finally, the circumstances in which the work was created (usually only known to the author of the work) may influence the interpretation of an image. The most difficult and problematic task is to express the highest level of meaning, which is the third level of Panofsky/Shatford model: iconology/interpretation. Only specialists or very experienced indexers will recognize some signs and symbols and, by reading between the lines, reconstruct the highly sophisticated iconographical program, e.g. “an image of a grey brick wall is about socialism” (Neugebauer, 2005, p.5). As Herbert Read noted “What we really expect in a work of art is a certain personal element – we expect artist to have, if not a distinguished mind, at least a distinguished sensibility. We expect him to reveal something to us that is original – a unique and private vision of the world.” (Read, 1968, p. 27) Unfortunately, very often we are unable to name and express what is important, unique and original in this vision, its sense remaining inaccessible or unexpressed.
In addition to the general problems with indexing of images, there are some questions connected with the rules governing the creation of the authorized access points and with the usage of the subject authority file. In this context a few specific problems can be enumerated:
- usage of detailed description vs. assignment of broader concepts,
- relation with authority file and with general indexing policy - consistency or independence?
- descriptive cataloguing and subject indexing – complementarity or independence?
Detailed description vs. broader concepts
It is obvious that the first and second level of images content should be indexed, but each cataloguing institution has to make a choice between assignments of strictly adequate or broader concepts. It is a crucial decision and not an easy one.
For example, Caravaggio’s Basket of Fruits may be indexed using single term referring to the art genre – “Still life”. For advanced iconographic researches, a detailed description (“apple; fig; grape; pear; basket”) could be more useful. A print which presents over twenty portraits of famous Italian singers may be indexed with subject heading “Singers” (with geographical subdivision “-Italy”) and general subject designation “Portrait”. Is it enough? The other option is to enumerate, as additional subject access points, all the names of the fifty depicted person: “Agujari, Lucrezia (ca 1743-1783); Amicis, Anna Lucia de (ca 1733-1816); Amorevoli, Angel (1716-1798); Ansani, Giovanni (1750-1815)” etc. Which of these solutions is better? Which of them could satisfy the need of various audiences and maximize research results?