Displaying data in multidimensional relevance space with 2D visualization maps

This paper introduces a tool for visualizing a multidimensi onal rel- evance space. Abstractly, the information to be displayed c onsists of a large number of objects, a set of features that are likely to be of interest to the user, and some function that measures the r ele- vance level of every object to the various features. The goal is to provide the user with a concise and comprehensible visualization of that information. For the type of applications we concentrate on, the exact rele- vance measures of the objects are not significant. This enabl es ac- curacy to be traded for a clearer display. The idea is to "flatt en" the multidimensionality of the feature space into a 2D "relevance map", capturing the inter-relations among the features, without causing too many ambiguous interpretations of the results. To bette r reflect the nature of the data and to resolve the ambiguity we refine th e given set of features and introduce the notion of composed features. The layout of the map is then obtained by grading it according to a set of rules and using a simulated annealing algorithm which opti- mizes the layout with respect to these rules. The technique we propose here has been implemented and tested, in the context of visualizing the result of a Web sear ch, in the RMAP (Relevance Map) prototype system.