An interactive, structure-mediated approach to exploring information in a heterogeneous, distributed environment
Terry Winograd, Michelle Q. Wang Baldonado · Medical Entomology and Zoology · 1997
Information exploration involves building bridges between document space (the realm of information artifacts, ranging from books to e-mail to bibliographic citations) and concept space (the context and knowledge required to interpret documents). This dissertation proposes an interactive, structure-mediated approach to exploring information in a heterogeneous, distributed environment--such as a digital library. It presents a conceptual model for information exploration, a system (SenseMaker) that embodies the proposed model, and several evaluative user studies. Two principles form the crux of our structure-mediated conceptual model, which supports users in moving back and forth between document space and concept space: (1) Users should be able to build document-based structures for document collections by interactively organizing them around contextual dimensions (e.g., author). These higher-level structures reveal key concepts (e.g., specific authors). (2) Users should be able to create new collections of documents by sculpting or adding to document-based structures. This reuse of structure enables concept-based filtering and concept-based finding of documents. This dissertation introduces iterative bundling and flexible duplicate detection as two strategies for interactively building document-based structure. Iterative bundling creates aggregates of similar documents. It allows the definition of similarity to vary according to contextual dimension. For example, a different metric might be used for bundling by author than for bundling by topic. Flexible duplicate detection identifies equivalent documents and merges them together. Again, the definition of equivalence can vary. The creation of a document collection is a prerequisite to building document-based structure for a collection. This dissertation defines an ontology of collection-creating actions that includes specifying the characteristics of a new collection, expanding a previous collection, limiting a previous collection, and returning to a previous collection. Collection-creating actions that utilize structure fit into several of these categories. The SenseMaker system embodies the proposed structure-mediated conceptual model. SenseMaker utilizes protocols and other technologies developed for the Stanford Digital Library project. Its interface design introduces the concept of hi-cites, a user-interface construct to facilitate browsing.