We present ThemeDelta, a visual analytics system for extracting and visualizing temporal trends, clustering, and reorganization in time-indexed textual datasets. ThemeDelta is supported by a dynamic temporal segmentation algorithm that integrates with topic modeling algorithms to identify change points where significant shifts in topics occur. This algorithm detects not only the clustering and associations of keywords in a time period, but also their convergence into topics (groups of keywords) that may later diverge into new groups. The visual representation of ThemeDelta uses sinuous, variable-width lines to show this evolution on a timeline, utilizing color for categories, and line width for keyword strength. We demonstrate how interaction with ThemeDelta helps capture the rise and fall of topics by analyzing archives of historical newspapers, of U.S. presidential campaign speeches, and of social messages collected through iNeighbors, a web-based social website. ThemeDelta is evaluated using a qualitative expert user study involving three researchers from rhetoric and history using the historical newspapers corpus.

Abstract Visual exploration involves using visual representations to investigate data where the goals of the process are unclear and poorly defined. However, this often places unduly high cognitive load on the user, particularly in terms of keeping track of multiple investigative branches, remembering earlier results, and correlating between different views. We propose a new methodology for automatically spatializing the individual steps in visual exploration onto a large visual canvas, allowing users to easily recall, reflect, and assess their progress. We also present a web‐based implementation of our methodology called ExPlatesJS where users can manipulate multidimensional data in their browsers, automatically building visual queries as they explore the data.

Many 2D visual spaces have a virtually one-dimensional nature with very high aspect ratio between the dimensions: examples include time-series data, multimedia data such as sound or video, text documents, and bipartite graphs. Common among these is that the space can become very large, e.g., temperature measurements could span a long time period, surveillance video could cover entire days or weeks, and documents can have thousands of pages. Many analysis tasks for such spaces require several foci while retaining context and distance awareness. In this extended version of our IEEE PacificVis 2010 paper, we introduce a method for supporting this kind of multifocus interaction that we call stack zooming. The approach is based on building hierarchies of 1D strips stacked on top of each other, where each subsequent stack represents a higher zoom level, and sibling strips represent branches in the exploration. Correlation graphics show the relation between stacks and strips of different levels, providing context and distance awareness for the foci. The zoom hierarchies can also be used as graphical histories and for communicating insights to stakeholders and can be further extended with annotation and integrated statistics.

We propose the notion of composite visualization views (CVVs) as a theoretical model that unifies the existing coordinated multiple views (CMV) paradigm with other strategies for combining visual representations in the same geometrical space. We identify five such strategies - called CVV design patterns - based on an extensive review of the literature in composite visualization. We go on to show how these design patterns can all be expressed in terms of a design space describing the correlation between two visualizations in terms of spatial mapping as well as the data relationships between items in the visualizations. We also discuss how to use this design space to suggest potential directions for future research.

We present gravity navigation (GravNav), a family of multi-scale navigation techniques that use a gravity-inspired model for assisting navigation in large visual 2D spaces based on the interest and salience of visual objects in the space. GravNav is an instance of topology-aware navigation, which makes use of the structure of the visual space to aid navigation. We have performed a controlled study comparing GravNav to standard zoom and pan navigation, with and without variable-rate zoom control. Our results show a significant improvement for GravNav over standard navigation, particularly when coupled with variable-rate zoom. We also report findings on user behavior in multi-scale navigation.

The most common techniques for navigating in multiscale visual spaces are pan, zoom, and bird's eye views. However, these techniques are often tedious and cumbersome to use, especially when objects of interest are located far apart. We present the PolyZoom technique where users progressively build hierarchies of focus regions, stacked on each other such that each subsequent level shows a higher magnification. Correlation graphics show the relation between parent and child viewports in the hierarchy. To validate the new technique, we compare it to standard navigation techniques in two user studies, one on multiscale visual search and the other on multifocus interaction. Results show that PolyZoom performs better than current standard techniques.

Time is a universal and essential aspect of data in any investigative analysis. It helps analysts establish causality, build storylines from evidence, and reject infeasible hypotheses. For this reason, many investigative analysis tools provide visual representations designed for making sense of temporal data. However, the field of visual analytics still needs more evidence explaining how temporal visualization actually aids the analysis process, as well as design recommendations for how to build these visualizations. To fill this gap, we conducted an insight-based qualitative study to investigate the influence of temporal visualization on investigative analysis. We found that visualizing temporal information helped participants externalize chains of events. Another contribution of our work is the lightweight evaluation approach used to collect, visualize, and analyze insight.

Animated transitions are popular in many visual applications but they can be difficult to follow, especially when many objects move at the same time. One informal design guideline for creating effective animated transitions has long been the use of slow-in/slow-out pacing, but no empirical data exist to support this practice. We remedy this by studying object tracking performance under different conditions of temporal distortion, i.e., constant speed transitions, slow-in/slow-out, fast-in/fast-out, and an adaptive technique that slows down the visually complex parts of the animation. Slow-in/slow-out outperformed other techniques, but we saw technique differences depending on the type of visual transition.

Direct manipulation has had major influence on interface design since it was proposed by Shneiderman in 1982. Although directness generally benefits users, direct manipulation also has weaknesses. In some cases, such as when a user needs to manipulate small, attribute-rich objects or multiple objects simultaneously, indirect manipulation may be more efficient at the cost of directness or intuitiveness of the interaction. Several techniques have been developed over the years to address these issues, but these are all isolated and limited efforts with no coherent underlying principle. We propose the notion of Surrogate Interaction that ties together a large subset of these techniques through the use of a surrogate object that allow users to interact with the surrogate instead of the domain object. We believe that formalizing this family of interaction techniques will provide an additional and powerful interface design alternative for interaction designers, as well as uncover opportunities for future research.

We evaluate unguided and guided visual search performance for a set of techniques that mitigate occlusion between physical and virtual objects on a tabletop display. The techniques are derived from a general model of hybrid physical/virtual occlusion, and take increasingly drastic measures to make the user aware of, identify, and access hidden objects—but with increasingly space-consuming and disruptive impact on the display. Performance is different depending on the visual display, suggesting a tradeoff between management strength and visual space deformation. These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Information visualization shows tremendous potential for helping both expert and casual users alike make sense of temporal data, but current time series visualization tools provide poor support for comparing several foci in a temporal dataset while retaining context and distance awareness. We introduce a method for supporting this kind of multi-focus interaction that we call stack zooming. The approach is based on the user interactively building hierarchies of 1D strips stacked on top of each other, where each subsequent stack represents a higher zoom level, and sibling strips represent branches in the visual exploration. Correlation graphics show the relation between stacks and strips of different levels, providing context and distance awareness among the focus points. The zoom hierarchies can also be used as graphical histories and for communicating insights to stakeholders. We also discuss how visual spaces that support stack zooming can be extended with annotation and local statistics computations that fit the hierarchical stacking metaphor.

Line graphs have been the visualization of choice for temporal data ever since the days of William Playfair (1759-1823), but realistic temporal analysis tasks often include multiple simultaneous time series. In this work, we explore user performance for comparison, slope, and discrimination tasks for different line graph techniques involving multiple time series. Our results show that techniques that create separate charts for each time series--such as small multiples and horizon graphs--are generally more efficient for comparisons across time series with a large visual span. On the other hand, shared-space techniques--like standard line graphs--are typically more efficient for comparisons over smaller visual spans where the impact of overlap and clutter is reduced.