Do you have large amounts of data but struggle to make sense of them?

Still generating thousands of visualizations manually and browsing through each one at a time? 

What if you could directly skip through this tedious process and get to the insights in your data directly? 


 Zenvisage fast-forwards users to desired insights by automatically identifying and recommending interesting visualizations. 

You specify at a high-level what you are looking for, and the system does the rest!


Get Started   Live Demo   Documentation

Key Features

Pattern Sketching

In Zenvisage, you can look for visualizations of interest by sketching the desired pattern on the querying canvas. For example, searching for cities with housing prices that rises at the beginning of the year then drops in the later part of the year.

Drag-and-Drop

You can also search for visualizations by dragging and dropping a visualization into the query canvas. For example, you can look for other cities that have a similar housing pattern to New York City.

Rank and Browse

Given a queried pattern, Zenvisage displays a ranked list of visualizations based on how similar the visualization is to the queried pattern. We also recommend representative trends and outliers to highlight the different types of patterns in the dataset.

Real-World Use Cases

Zenvisage has been developed in close collaboration with a number of users who have large datasets and a pressing need for automated data exploration. Our participatory design process for these use cases is detailed in this paper. Examples include:

Material Science Battery Discovery

Using dynamic classes, scientists can compare characteristics from different data classes to find a solvent datapoint that satisfies desirable properties, such as high-voltage or low solubility. By browsing through patterns in different data classes, they can also study the overall tradeoffs and relationships between data attributes.

Astronomical Transient Discovery

By sketching a peak-then-decay pattern on the query canvas, astronomers can directly search for potential supernovae candidates. Using interactive filtering, they can also eliminate sources of data anomaly to improve match accuracy for finding candidates.

Gene Expression Analysis

Using the recommended representative trends in Zenvisage, geneticists can learn about the characteristic pattern profiles in their dataset. They can also search and browse for genes belonging to the same representative cluster via drag-and-drop.

Zenvisage has been used by members of the following institutions. If you have a potential use case for Zenvisage or are currently using it for your project, we'd love to hear from you!

Demo

This demo introduces functionalities from our latest software release (v3.0) for Zenvisage, including different ways for constructing visual queries to search for patterns of interest, as well as additional features, such as interactive data filtering, dynamic class creation, data smoothing.

Try out our live demo !

Contact Us

Zenvisage is being developed by a team of undergraduate and graduate students headed by Professor Aditya Parameswaran and Professor Karrie Karahalios. The current list of contributors includes (in alphabetical order) Doris Lee, Jaewoo Kim, Jintao Jiang, and Renxuan Wang. Past contributors include (in alphabetical order) Lijin Guo, John Lee, Changfeng Liu, Albert Kim, Tarique Siddiqui, Chao Wang, Edward Xue, Sean Zou, and Yuxuan Zou. There are other students working on related projects: Silu Huang, Stephen Macke, Sajjadur Rahman, and Tana Wattanawaroon.
Please reach out to Doris Lee if you'd like to either contribute or be a beta tester of Zenvisage!

Publications

© Zenvisage 2019. With generous funding from Google, Siebel Energy Institute, and Toyota.