Visualizing a Knowledge Domain's Intellectual Structure

V isualizing the entire body of scientific knowledge and tracking the latest developments in science and technology have intrigued generations of scientists, philosophers, government officials, librarians, and publishers. Advances in information visualization offer promising tools for presenting knowledge structures and their development in an increasingly intuitive way. 1 The scientific literature provides ingredients ripe for knowledge visualization. Researchers commonly focus on significant structural patterns in knowledge discovery , information retrieval, and other disciplines that may offer insights into the nature of underlying interrelationships such as those among authors of scholarly publications, 2 documents, 3 and journals. 4 We describe an approach to visualizing a knowledge domain's intellectual structures that extracts structural patterns from the scientific literature and represents them in a 3D knowledge landscape. This approach extends and transforms traditional author co-citation analysis (ACA) into a knowledge-visualization and domain-analysis tool. Several existing systems address common knowledge-visualization issues, such as selecting appropriate similarity metrics and displaying high-dimensional structures. SemNet, introduced in the 1980s, produces 3D graphic representations of large knowledge bases to help users grasp complex relationships. 5 SemNet's design focuses on the graphical representations of three types of components: • the identity of individual elements in a large knowledge base, • the relative position of an element within a network context, and • explicit relationships between elements. SemNet represents elements of a Prolog rules knowledge base as labeled rectangles connected by lines or color-coded arcs. A Prolog module, which contains a subset of the logic programming language's rules, thus appears as a rectangle labeled with the module's name. In SemNet, the closeness between two rectangles indicates the strength of the connection between their modules. To show how the knowledge base works, SemNet's designers experimented with various techniques such as multidimensional scaling (MDS), simulated annealing, fisheye views, and even a sprite that travels down arcs between rectangles. The Spatial Paradigm for Information Retrieval and Exploration, developed by Pacific Northwest National Laboratory, provides a classic example of information visualization. 3 Spire consists of a suite of visualization tools for browsing large sets of documents—also called a document collection or corpus—and includes a well-known visualization view called Theme-scape. This tool creates an abstract, 3D landscape view of a document corpus. A thematic terrain simultaneously communicates both the primary themes of the underlying document collection and a measure of their relative prevalence in the collection. Thematic peaks and valleys in Themescape produce a simplified …