Organizing moving image collections for the digital era: the goldspeil report

Information Outlook, August, 2002 by James M. Turner, Michele Hudon, Yves Devin

The rapid buildup of collections and the need to communicate between systems via networks add urgency to the need for development of common methods for storage and retrieval. One important contributor to communication would be a common thesaurus for shot-level indexing.

Analysis and Representation of Moving Images

Images can be analyzed and interpreted in diverse ways. One model widely used in image indexing, based on the work of the art historian Erwin Panofsky (1955), identifies three levels of interpretation. The first, which Panofsky calls pre-iconographic, deals with the primary or "natural" subject of an image. The second, iconographic, has to do with secondary or conventional subject matter. The third, iconologic, deals with symbolic levels of interpretation. Shatford (1986) emphasizes the first and second levels, and translates them into the "ofness" and the "aboutness" of a picture. What is it a picture of? What is the picture about? These levels also correspond rather closely to the ideas of denotation and connotation in the area of semiotics.

Whether they are still or moving, pictures contain a great variety of information and can have different meanings for different viewers (Shatford 1986, 42). This fact can of course be a source of problems in working out uniform methods for describing pictures for purposes of storage and retrieval. For stockshots, it has been suggested that only the primary level is really useful (Turner 1990, 12). Most moving image collections described at the shot level are indexed this way, and the descriptors necessary for representing the visual content simply name the objects, persons and events found in the shots (e.g., a cat or a chair) rather than abstract notions (e.g., comfort or serenity). From this perspective, using a thesaurus as an indexing tool is of great interest. The lexical and structural control such a tool offers can contribute greatly to improving access to the content of collections, to reducing noise (i.e., too many hits) and silence (i.e., too few hits) in retrieval, to improving precision and ultima tely to satisfying users by giving them what they need without requiring a great investment of their time.

The development of thesauri is based on rules and principles spelled out in international norms (Hudon 1994, 75-76). Thesauri are dynamic tools that are adaptable to new realities and new needs of the collections that use them. The content can constantly be updated and improved to meet the needs of users, especially when database software is used for managing the thesaurus. However, this tool is relatively exclusive, and is usually developed to manage the vocabulary of a particular area of endeavor and that of a particular group of users (Van Slype 1987, 117). Unlike classification systems and lists of subject headings, thesauri are not encyclopedic. Yet such a thesaurus, general yet encompassing, is what is needed for the shot-level description of moving image collections.

A few thesauri have been created specifically for indexing visual documents such as art images, photos, slides and plans. Perhaps the best known of these is the Art and Architecture Thesaurus (2001) managed by the Getty Foundation. The AAT offers a standardized terminology of about 40,000 expressions covering art and architecture from antiquity to the present day. The Thesaurus for Graphic Materials (2001), published by the Library of Congress, offers descriptors for indexing printed pictures, photos, drawings, cartoons, posters and architectural drawings. In Canada, the National Film Board's thesaurus for indexing stockshots is also of use. In addition, there are a number of visual thesauri available. These take the approach of using pictures instead of text to represent other pictures (Rasmussen 1997, 182), offering a clear advantage in a multilingual environment. The NASA Visual Thesaurus is an example of this type of tool.