Reassembling Elephants: A Multi-Spatiotemporal Visualization Method for History and Humanities Data

In the wake of digitization initiatives, we find a wide range of phenomena in historical fields of study translated into complex, event-based data. From large-scale accounts on collective developments (such as cultural histories, art-historical movements, or socio-technical translation processes) to individual biographies of actors and objects - datafication initiatives frequently utilize event-based data formats to model complex topics as constellations of historical data points. Aside from time stamps, these data points then often are defined by further attributes, which provide their geographic, categorical, or relational specifics. The availability of such complex datasets opens new opportunities for DH research and teaching, but also for the public communication of humanities topics, and for open science endeavors.However, when engaging in the visual analysis and communication of complex historical data, scholarly or public audiences rarely get to see their multidimensional richness. Commonly, visualization tools require analysts to selectively ‘cut’ into the complexity of the data to highlight and project particular aspects, while neglecting other facets and data dimensions. While coordinated multiple views allow us to establish plurality of perspectives side by side (Dörk et al., 2018; Roberts et al., 2019), they come with a variety of downsides on their own, including the splitting of users’ attention, and a significant amount of visual complexity (Baldonado et al., 2000; Windhager et al., 2018a). Figuratively speaking, multiple views allow us to grasp, see, and sample vital parts of the proverbial elephant, while hindering us to see the whole, dynamic organism in its particular context. If we want to overcome this state of affairs, we have to (re)connect and (re)assemble the partial impressions from multiple views for ourselves, which turns out to be a demanding cognitive task (Windhager et al., 2018a).To provide a more integrated approach to the analysis of event-based data, we introduce the PolyCube visualization framework. As a web-based visualization system, it draws together multiple perspectives to convey a bigger picture for complex, time-oriented data, and to support synoptic exploration of the data, as well as navigation between specific perspectives for expert and casual users alike. The system provides:multiple spatialized (i.e. geographic and non-geographic) overview perspectives (including a map-based, a set-based, and a network-based view),multiple perspectives on the temporal data dimension (including space-time cuberepresentation, juxtaposition, animation and superimposition views)close-up access to single events or objects on demand, together withanimated canvas transitions, supporting the switching between various views.To demonstrate the analytical options of this system, we turn to the field of cultural heritage studies, and showcase the visual analysis of two cultural heritage collections (Windhager et al., 2020), including the Charles W. Cushman photography collection (Indiana University, 2007), and a corpus of influential movies, based on cinematic