Segment, enrich, annotate, and link (SEAL) paradigm for context-aware access to complex and structured data.

摘要

Creation and access to information are becoming integral parts of our daily activities. Many aspects of our lives, such as education, work, commerce, and science and research, are increasingly data and information heavy; many of our tasks require data from diverse sources and produce information to be recorded for later use. Yet, it is also becoming more and more difficult for individuals to navigate through complex and increasingly growing information spaces to locate information relevant in a particular context. Therefore, there is an impending need for effective mechanisms that will reduce an individuals' search and navigation overhead within complex and dynamically growing information spaces.;This thesis introduces a segment-enrich-annotate-link (SEAL) paradigm to support effective navigation and search in complex and structured information space. In particular, by leveraging the SEAL paradigm, this work investigates and builds efficient and effective frameworks for improved information access within various different application domains, including the design of accessible educational information spaces for individuals with visual disabilities; reuse-based organization and tagging of entries in blogosphere; ontology-supported, context-aware search and navigation within digital libraries; and relationship mining within complex scientific databases. The SEAL paradigm is instantiated and evaluated through various applications, including the iCare-Assistant (an assistive system for helping students who are blind in accessing web and electronic course materials) and the tDAR (a digital library which provides integrated access to archaeological data collected by independent research groups). Evaluations show that these systems built based on the SEAL paradigm reduces the navigational overhead significantly and enables user to access complex and structured data effectively.;Most of data in various applications are organized in a structure. In structured data, the relationship between pairs of data nodes reflects the underlying semantic properties (i.e., specialization/generalization). While structures help organize information units, they also provide challenges since the data nodes cannot be considered alone, but should be considered with the semantic context implied by the underlying structure. Thus, any mechanism that helps users navigate in a structured information space necessitates understanding of the structural and semantic relationship implied by the data nodes in structured data.