This work proposes the extension of a framework for visual mining of text collections developed in the last three years within a larger information visualization project. The approach adopted thus far includes, among other tools, building visualizations that map text data sets onto a 2D plane using multidimensional or distance based projections or point placement via similarity trees. A basic system already exists that offers a complete set of tools for building such maps. However it is almost universely agreed in the visualization community that only having multiple visions of a certain data set is not enough to allow full exploration of complex data sets. It is necessary to have a consistent means to coordinate such views. In the case of text mappings there is very little done in terms of coordinating visualizations, particularly where mining strategies are involved (for instace, for topic extraction) as in our case. This project targets at the development of a methodology for coordination between multiples views of document maps in the context of visual mining, its implementation in the framework currently under development (the Projection Explorer - PEx), and its evaluation concerning a number of data sets and general users.
News published in Agência FAPESP Newsletter about the scholarship: