Visualising document content with metadata to facilitate goal-directed search

Much corporate knowledge is captured in documents. This paper describes the design and evaluation of an application, named Gridvis, which uses metadata to enable rapid goal-directed searches within corporate documents. Gridvis visualises metadata that describes and differentiates the contents of individual paragraphs. This visualisation is used as the basis for rapid querying and navigation. An experimental evaluation of Gridvis´s performance on representative work tasks is described. Qualitative analysis of the data suggests that there is substantial room for performance improvement, and has inspired design changes to realise this potential. Quantitative analysis proved inadequate to detect a numerical advantage due to the sample size restrictions inherent when doing academic research in a corporate setting. The problems of empirical evaluation in a corporate setting are discussed. An alternative approach to reliable evaluation is proposed