| Purpose. This study provides a preliminary scientific account of indexing expertise from a cognitive perspective and suggestions for indexer training and the design of computer-assisted indexing systems. It explores the cognitive processes of six indexers performing document-oriented indexing of empirical journal articles in the domain of agriculture.; Method. This study used a qualitative approach. Six indexers from the Indexing Branch of the National Agricultural Library participated. Each of six documents was indexed by three indexers differing in domain expertise and indexing experience in the subject area of the document. Four methods were used for data collection: thinking aloud, observation, document analysis, and interviewing. The resulting textual data were analyzed with the aid of NUDIST, a qualitative research tool.; Findings. Three overall approaches to generating a list of indexing descriptors were identified--text-as-a-whole, text-segment-with-notes, and text-segment-directly-to-descriptors. A complex pattern of component processes emerged: attending to different portions of the text of a document with varying intensity; interpreting the subject contents of the text (recognizing the general theme, using frames and semantic clusters of concepts for identifying indexable concepts, organizing indexable concepts into an integrated framework, and drawing inferences with respect to implications); judging indexability of concepts and deciding on the specificity of indexing; transforming indexable concepts to index terms following semantic paths; and self-evaluating the indexing. The indexers relied on both internal and external information sources. Among indexer variables, most important is the influence of domain expertise on the semantic level of information processing. Document variables, such as length, flow of thoughts, methodological significance, and single- vs. multi-topics also had an effect.; Implications. We present suggestions for designing improved computer-assisted indexing systems at two levels: low-level assistance (accessing external information sources, searching the text for occurrences of a word, retrieving similar records, and providing a voice input electronic note pad) and knowledge-base assistance (for example, scanning the text for words that could serve as values in a frame). All the findings can be used in designing training programs for indexers. The findings on semantic paths can be used to derive a richer set of semantic associations among terms in thesauri. |
