SE-3D: A controlled comparative usability study of a virtual reality semantic hierarchy explorer.

摘要

Keyword searching (e.g., Google or Yahoo!) is based on uncontrolled vocabulary matching which often produces large and noisy result sets. This can waste the time of the searcher who has to sift through long lists of often irrelevant information. The Semantic Web initiative aims to address this issue and includes the description of content using controlled ontologies (i.e., sets of descriptive terms and their relations). Ontologies are partly hierarchical structures too large to display on a single computer screen and thus difficult for searchers to explore efficiently.;SE-3D was tested by 24 undergraduate students during a repeated measures within-subject experiment. As compared with a text-only baseline, SE-3D produced an advantage in accuracy. Participants were more patient with SE-3D, they preferred it and perceived it as more useful. The application used a new technique to manage hundreds of overlapping textual labels in virtual reality, and offered a novel integration of explorative and specific keyword searching. The analysis of the collection revealed that subject assignments followed a power law; the top 1% most assigned subjects contained over 58% of the collection and 65% of non-empty subjects contained a single document.;The findings suggest it is possible to extract additional value from organized collections by offering untrained users a reconstructed subject structure integrated with keyword searching. This research is significant for the development and testing of improved bridges between information organization and IR, and interactive information visualization.;In an attempt to address these issues, this research has developed and tested Subject Explorer 3D (SE-3D): an information visualization (IV) virtual reality (VR) information retrieval (IR) application based on the metaphor of exploring a physical space. SE-3D aimed to facilitate the visual exploration of information by offering searchers an interactive representation of the subject structure found in the Library of Congress Subject Headings (LCSH). SE-3D is a visual subject ontology navigation tool integrated with keyword searching and relevance ranking of a realworld information collection.