In the "big data" era, everyone has access to massive quantities of data, and are struggling to
make sense of, and derive value from such data. The state of the art for non-programmers
is to load this data into a
visualization tool, and repeatedly generate visualizations until desired ones are identified.
This exploration is painful, tedious, and time-consuming, meaning that the "insight per unit
time" is exceedingly low.
What if you could "skip ahead to the insights"?
Zenvisage is a visual exploration system that can automatically
identify and recommend interesting
visualizations. The user can specify at a high level what they are looking for, and the system will
do the rest.
Zenvisage enables users to effortlessly receive visualization recommendations for interesting trends, patterns, and insights from large datasets. Here are the key features of Zenvisage:
Zenvisage generalizes and extends our prior system SeeDB (see papers).
- Interactive Insight Sketch Interface.
Zenvisage users can directly draw the trend-line, bar chart, or scatterplot they are looking for, and then rely on the system to find appropriate matches: for instance, a person browsing a dataset of material properties may be looking for those materials displaying a specific correlation between two properties. Users can also drag-and-drop trends onto the canvas and then subsequently modify the trend.
Using this interface, users can specify the insights they are looking for, and expect Zenvisage to find matches, much like a "visualization search engine".
- Sophisticated Visual Query Interface.
For more complex requests, Zenvisage supports a query language, called ZQL (pronounced "zee-quel"), short for Zenvisage Query Language, a flexible and intuitive mechanism to specify desired insights from visualizations. Using a small number of ZQL lines, users can explore trends, patterns, and insights in any which way they desire.
- Visualization Recommendations. In addition to returning results for user-submitted queries, zenvisage runs a host of parallel queries to find the most interesting trends for that subset of data the user is currently viewing and presents them as visualization recommendations.