Exact versus Estimated Pruning of Subject Hierarchies

摘要

Many large digital collections are organized by subject;these useful information organization structures are largeand complex, thus difficult to browse. Current online toolsand visualization prototypes show small localized subsetsand do not provide the ability to explore the predominantpatterns of the overall subject structure. This study buildson existing work concerning automatic subject hierarchymodification techniques that aim to facilitate browsing fordocuments by capitalizing on the highly uneven distributionof real-world collections. Specifically, previous work usedan estimation of the number of accessible documentsoffered by each subject term, while the current study usesthe exact number of accessible documents. The impact isdemonstrated on a large collection organized using MedicalSubject Headings (MeSH). Results show that, althoughcomputationally more demanding, pruning the MeSHhierarchy based on the exact access produces a differentsubject hierarchy under some conditions. The visual impactis demonstrated using traditional outline views. This studyhas implications for the development of informationorganization theory and human-information interactiontechniques for subject hierarchies.