Visualization of Multi-domain Ranked Data

This chapter focuses on the visualization of multi-domain search results. We start by positioning the problem in the recent line of evolution of search engine interfaces, which more and more are capable of mining semantic concepts and associations from text data and presenting them in sophisticated ways that depend on the type of the extracted data. The approach to visualization proposed in search computing extends current practices in several ways: the data to visualize are N-dimensional combinations of objects, with ranking criteria associated both to individual objects and to sets of combinations; object's properties can be classified in several types, for which optimized visualization families are preferred (e.g., timelines for temporal data, maps for geo-located information); combinations may exhibit any number of relevant properties to be displayed, which need to fit to the hi-dimensional presentation space, by emphasizing the most important attributes and de-emphasizing or hiding the less important ones. The visualization problem therefore amounts to deciding the best mapping between the data of the result set and the visualization space.