論文アブストラクト： Analysts need interactive speed for exploratory analysis, but big data systems are often slow. With sampling, data systems can produce approximate answers fast enough for exploratory visualization, at the cost of accuracy and trust. We propose optimistic visualization, which approaches these issues from a user experience perspective. This method lets analysts explore approximate results interactively, and provides a way to detect and recover from errors later. Pangloss implements these ideas. We discuss design issues raised by optimistic visualization systems. We test this concept with five expert visualizers in a laboratory study and three case studies at Microsoft. Analysts reported that they felt more confident in their results, and used optimistic visualization to check that their preliminary results were correct.
論文アブストラクト： Interactive exploration plays a critical role in large graph visualization. Existing techniques, such as zoom-and-pan on a 2D plane and hyperbolic browser facilitate large graph exploration by showing both the details of a focal area and its surrounding context that guides the exploration process. However, existing techniques for large graph exploration are limited in either providing too little context or presenting graphs with too much distortion. In this paper, we propose a novel focus+context technique, iSphere, to address the limitation. iSphere maps a large graph onto a Riemann Sphere that better preserves graph structures and shows greater context information. We conduct extensive experiment studies on different graph exploration tasks under various conditions. The results show that iSphere performs the best in task completion time compared to the baseline techniques in link and path exploration tasks. This research also contributes to understanding large graph exploration on small screens.
論文アブストラクト： We present TagRefinery, an interactive visual application aiding the cleaning and processing of open tag spaces, such as those in Last.fm or YouTube. Our pre-design analysis showed a need to support a spectrum of user expertise from novice to advanced, which resulted in two distinct interface modes. Summative evaluations of TagRefinery showed that it could effectively guide the novice users through the workflow by giving them brief but helpful explanations on why each step was required, and providing visual and statistical aids to help them in making important decisions. This is while our more expert users greatly appreciated the amount of control and granularity over the workflow that our more advanced interface mode offered. Both the underlying tag cleaning workflow and the interface were designed iteratively in a participatory design process in collaboration with research on a music recommendation interface based on Last.fm tags.
論文アブストラクト： Spatial datasets, such as tweets in a geographic area, often exhibit different distribution patterns at multiple levels of scale, such as live updates about events occurring in very specific locations on the map. Navigating in such multi-scale data-rich spaces is often inefficient, requires users to choose between overview or detail information, and does not support identifying spatial patterns at varying scales. In this paper, we propose TopoGroups, a novel context-preserving technique that aggregates spatial data into hierarchical clusters to improve exploration and navigation at multiple spatial scales. The technique uses a boundary distortion algorithm to minimize the visual clutter caused by overlapping aggregates. Our user study explores multiple visual encoding strategies for TopoGroups including color, transparency, shading, and shapes in order to convey the hierarchical and statistical information of the geographical aggregates at different scales.