HMaViz: Human-machine analytics for visual recommendation

Visualizations are context-specific. Understanding the context of visualizations before deciding to use them is a daunting task since users have various backgrounds, and there are thousands of available visual representations (and their variances). To this end, this paper proposes a visual analytics framework to achieve the following research goals: (1) to automatically generate a number of suitable representations for visualizing the input data and present it to users as a catalog of visualizations with different levels of abstractions and data characteristics on one/two/multi-dimensional spaces (2) to infer aspects of the user's interest based on their interactions (3) to narrow down a smaller set of visualizations that suit users analysis intention. The results of this process give our analytics system the means to better understand the user's analysis process and enable it to better provide timely recommendations.