Using Context to Assist in Personal File Retrieval
Carnegie Mellon University School of Computer Science Ph.D. Dissertation CMU-CS-06-147, August 2006.
Craig A.N. Soules
School of Computer Science
Carnegie Mellon University
Pittsburgh, PA 15213
Personal data is growing at ever increasing rates, fueled by a growing market for personal computing solutions and dramatic growth of available storage space on these platforms. Users, no longer limited in what they can store, are now faced with the problem of organizing their data such that they can find it again later. Unfortunately, as data sets grow the complexity of organizing these sets also grows. This problem has driven a sudden growth in search tools aimed at the personal computing space, designed to assist users in locating data within their disorganized file space.
Despite the sudden growth in this area, local file search tools are often inaccurate. These inaccuracies have been a long-standing problem for file data, as evidenced by the downfall of attribute-based naming systems that often relied on content analysis to provide meaningful attributes to files for automated organization.
While file search tools have lagged behind, search tools designed for the world wide web have found wide-spread acclaim. Interestingly, despite significant increases in non-textual data on the web (e.g., images, movies), web search tools continue to be effective. This is because the web contains key information that is currently unavailable within file systems: context. By capturing context information, e.g., the links describing how data on the web is inter-related, web search tools can significantly improve the quality of search over content analysis techniques alone.
This work describes Connections, a context-enhanced search tool that utilizes temporal locality among file accesses to provide inter-file relationships to the local file system. Once identified, these inter-file relationships provide context information, similar to that available in the world wide web. Connections leverages this context to improve the quality of file search results. Specifically, user studies with Connections see improvements in both precision and recall (i.e., fewer false-positives and false-negatives) over content-only search, and a live deployment found that users experienced reduced search time with Connections when compared to content-only search.
FULL THESIS: pdf