Subject analysis and indexing : from automated indexing to domain analysis

Preparing an index to a book or assigning indexing terms to documents involves the task of subject analy sis. Subject analysis can be defined as "the intellectual or automated process by which the subjects of a docu ment are analyzed for subsequent expression in the form of subject data'.1 Is automatic indexing a chal lenge, or even a threat to indexers? What do we mean when we talk about "subjects' of books and other documents? Are there different conceptions of subjects, and hence of subject analysis; and if so, are such con ceptions interconnected with methods applied for indexing? These questions are important because discussions of indexing tend to focus on the performance of auto mated indexing2 versus the performance of human indexing,' usually based on relevance judgments com piled from particular groups of IR-system users,4 or observations of interindexer consistency.Using perfor mance as a criterion for finding a 'winning' approach to subject analysis and indexing implies a model which restricts itself to comparing the two techniques. Based on empirical observations of user and indexer behav iour, it represents a mechanistic evaluation method for subject assignment and retrieval. This article presents an alternative model for discussing subject analysis and indexing, where the choice between particular methods applied for this task is of minor importance. The inten tion is to attempt to place indexing in a wider social context beyond such mechanistic evaluation methods and to point towards new challenges for us as indexers.