Enhancing retrieval and discovery of desktop documents

Personal computers provide users with abilities to create, organize, store and access large amounts of information. Most of this information is in the form of documents in files organized in the hierarchical folder structures provided by the operating system. Operating system-provided access to these data is mainly through structure-guided navigation, and more recently through keyword search.This thesis describes the author's research into the accessibility and utilization of personal documents stored and organized using the hierarchical file system provided by common operating systems. An investigation was carried out on how users currently store and access their documents in these structures. Access and utility problems triggered a need to reconsider the navigation methods currently provided. Further investigation into navigation of personal document hierarchies using semantic metadata derived from the documents was carried out. A more intuitive exploratory interface that exposes the metadata for browsing-style navigation was implemented. The underlying organization is based on a model for navigation whereby documents are represented using index terms and associations between them exposed to create a linked, similarity-based navigation structure. Exposure of metadata-derived index terms in an interface was hypothesized to reduce the user's cognitive load and enable efficient and effective retrieval while also providing cues for discovery and recognition of associations between documents. Evaluation results of the implementation supports this hypothesis for retrieval of deeply located documents, as well as better overall effectiveness in associating and discovery of documents. The importance of semantic document metadata is also highlighted in demonstrations involving transfer of documents from the desktop to other organized document stores such as a repository