Query Fingerprinting: Document Visualization Using Color Coding

摘要

This paper presents a document visualization methodology called Query Fingerprinting in which text-based information is transformed into an abstracted visual representation. The visualization process maps given queries to primary colors, and the frequency of each query in a paragraph defines the intensity of the corresponding query's color. The mixture of queries' colors is projected onto a sequence of strips that simultaneously show paragraph lengths, query distribution, and query relationships. This compact, but informative visual abstraction allows users to compare multiple documents intuitively and swiftly while alleviating the need to reference the underlying text. The Query Fingerprinting system also provides interactive zoom in functionality for further document navigation and for refinement of the search scope. An experimental study has been conducted on the Congressional text archive in which Query Fingerprinting has shown better performance in finding relevant information than the existing summary-based system, and subjects have shown a high degree of satisfaction using the Query Fingerprinting method.