Multi-representation Lens for Visual Analytics

Abstract

Modern data analysis deeply relies on computational visualization tools, specially when spatial data is involved. Important efforts in governmental and private agencies are looking for patterns and insights buried in dispersive, massive amounts of data (conventional, spatiotemporal, etc.). In Visual Analytics users must be empowered to analyze data from different perspectives, integrating, transforming, aggregating and deriving new representations of conventional as well as spatial data. However, a challenge for visual analysis tools is how to articulate such wide variety of data models and formats, specially when multiple representations of geographic elements are involved. A usual approach is to convert data to a database -e.g., a multirepresentation database -which centralizes and homogenizes them. This approach has restrictions when facing the dynamic and distributed model of the Web. In this paper we propose an on the fly and on demand multi-representation data integration and homogenization approach, named Lens, as an alternative that fits better with the Web. It combines a metamodel driven approach to transform data to a unifying multidimensional and multirepresentation model, with a middleware-based architecture for seamless and on-the-fly data access, tailored to Visual Analytics.