The Caldeson Picture Concept Indexing models

Introduction

Picture archive researchers know the problem: a client wants an image to illustrate an idea but hasn’t the least idea what would suit the concept. Answer: the trained, on-brain knowledge of the researcher about the contents of the archive and the business requirements of the client.

But, the answer is very much more difficult to find if the picture archive is on-line and the client has to use keyword or text search engines to track down the image.

It’s a problem facing many book, magazine and newspaper publishers as more and more big picture agencies digitise their collections, get their images on-line, let their armies of researchers go and expect clients to make their own searches remotely.

How can an automatic index help a client who wants an image to illustrate “hot”, for instance? The publisher wants to avoid the cliché flame or steam images. But, how does the search engine find the image of a sweating hand, a chilli pepper, a straining trumpeter, a surreal graphic or a tiger’s gaping gullet?

Without the help of an archive-wise researcher, how does a graphic designer find a novel image that says “dispirited”, “strong” or “slow” without descending to overworked depictions of, successively, mouth, muscle or mollusc?

The answer is concept indexing. Using keywords or text, archivists and/or collection filing staff have to register their images in a new thesaurus of words indicating concepts suggested by the picture, in addition to the indexer’s conventional Six States: “how, where, when, what, why and who”.

The Caldeson Consultancy’s Principal, Michael Steemson, has made a study of this intriguing problem and has identified three sectors for the concept indexing of images. He calls them “Actions”, “Concepts and Feelings” and “Facial Expressions”.

He explains: “Pictures have to be considered very differently from words when undertaking indexing. A single picture can illustrate many different moods or ideas, some of them totally dissimilar.

“Images are often ‘pictorially synonymous’, even though the events with which they are associated are antonymous – a grinning rioter, for instance, could become an image of humour or anger, a screaming pop fan indicate terror or delight, an explosion illustrate destruction or construction, a runner show flight or pursuit.”

During his study of the issue, he met one picture library researcher whose favourite image was of a glass-walled tower block viewed from street level up to blue sky scattered with white cloud. Different clients had used this single picture to illustrate “progress”, “excess” and “environmental pollution".

Action words

To illustrate action or inaction, for that matter, a thesaurus of some 40 words will cover the case, Michael Steemson found. Some of the 40 may have to be linked to synonyms -- smiling, leering, grinning, for instance, or dissenting and fighting -- but after a study of almost every verb in a dictionary, he found that pictorial action shots could be properly classified using just a 40-word thesaurus.

The 40 keywords are listed in the Action thesaurus model.

Facial Expressions

The human face is capable of expressing some 100 emotions, the study found, but in pictorial terms, these can be linked in groups of associated images to illustrate 16 basic human emotions like anger, defeat, compassion and so on.

For example, the images indexed with the words like apathy, boredom, complacency, contemplation, contentment, indifference, passivity, serenity, stupidity and vacant could all be used to illustrate the concept "peaceful". An even large number of words like anger, annoyance, contempt, dislike, evil, lust, madness, superiority or violence could be linked to illustrate "aggressive".

The 100 keywords in their 16 groups are listed in Facial Expressions thesaurus.

Concepts and feelings

This model comprises the biggest thesaurus and the largest number of linked words. It gathers some 500 moods and feelings into a list of key words covering concepts from “admiration” through “cold”, “ecstasy” and “hard” to “uselessness” and “work”.

In this model, particularly, images of similar moods can be grouped to indicate a single concept, images that may look the same but are described with words that are often not synonymous. For example, the concept of “ability” may be illustrated by such disparate key words as “achievement”, “craftsmanship”, “efficiency”, “self-control” and “superiority”. Similarly, the idea of “adversity” may come from suggestions of “difficulty”, “self-denial” or “vengeance”.

By grouping these “pictorially synonymous” words, almost 80 concepts, moods or feelings can be illustrated. The groups are listed in the Concepts and Feelings model.

Further study

For further study of indexing electronic archives, have a look at the the National Archives of Australia's document image digitisation project described and evaluated by Ted Ling, Director of the Archives' Legislative and Accessibility Projects, in his 2002 case study, Taking it to the streets: why Australia's National Archives has embraced digitisation on demand on-line at the Records and Information Management On-line Service (RIMOS).

Also, see the Thesaurus principles and practice paper on the Willpower Information, the British information management consultancy.

The Caldeson Consultancy Principal, Michael Steemson, would be glad to hear of any other studies of a similar kind, or other resources that may be of assistance to picture librarians.

Take a look at the Caldeson Guest Book. There might be useful comments there from previous visitors. To add a comment about this or any other Caldeson.com service, click HERE.

To go to the Caldeson Home Page, click HERE.

Michael Steemson

For more information, e-mail to: