SparkClouds

Information Visualization
Jamie Hand and Aayam Poudel

What is a SparkCloud?

A SparkCloud is a tag cloud with sparklines incorporated along with each word. It was a result of Microsoft Research conducted by Bongshin Lee, Nathalie Henry Riche, Amy K. Karlson, and Sheelagh Carpendale. The purpose of SparkClouds is to let people see how tag clouds change over time.

Our implementation

The source code for the implementation, including the Python scripts and SOTU text files used to generate our dataset, can be found on GitHub.

sparkcloud demo

Evaluation

SparkClouds, like all other visualizations, have pros and cons.

Pros

They let users compare multiple tag clouds without having to see them separately. They make it convenient to see how text samples change over time. SparkClouds make the visualization compact and less cluttered compared to alternatives. Scrolling is usually not necessary, and everything can be made to fit onto one screen. Identifying words is fast, and users can easily get a big picture of the data without interaction. Finally, color could also be used in the visualization to encode other things.

Cons

The lack of x-axis labels forces users to guess which time period (e.g. month) is represented. That is, accessing specific time and other details in the data would require interaction or something similar. The y-axis is normalized for each word, which makes it difficult to compare frequency between words (except by size of word, which is less precise). The position of a text on the screen means nothing, which means we're using the most powerful visual variable without it conveying anything meaningful. Furthermore, long words inevitably get more attention than short words of the same frequency, which might give a slightly wrong impression about the data. Finally, SparkClouds are a very specific type of visualization, which can only be used for word-based or tag-based datasets.

Summary

While SparkClouds are good for seeing trends in a tag-based dataset over time, they're not great for diving into details about the dataset. One particular example of when SparkClouds would be very useful is comparing State of the Union addresses over the years and seeing how the most frequently used words have changed over time. This gives us a big picture of how the speeches might have changed to reflect time era, presidents, and policies.

Presentation

Browse our presentation for background information, alternatives considered by the developers of SparkClouds, and pros and cons to the visualization method.

Works Cited

  1. Lee, Bongshin, Nathalie Henry Riche, Amy K. Karlson, and Sheelash Carpendale. "SparkClouds: Visualizing Trends in Tag Clouds." IEEE Trans. Visual. Comput. Graphics IEEE Transactions on Visualization and Computer Graphics 16.6 (2010): 1182-189. Web. http://research.microsoft.com/en-us/um/people/nath/docs/sparkclouds_infovis2010.pdf, Research Paper on SparkClouds
  2. https://en.wikipedia.org/wiki/Sparkline, Wiki Article on SparkLines
  3. https://commons.wikimedia.org/w/index.php?curid=14491713, WordCloud image
  4. http://research.microsoft.com/en-us/um/people/bongshin/, Bongshin Lee
  5. http://research.microsoft.com/en-us/um/people/nath/, Nathalie Henry Riche
  6. http://research.microsoft.com/en-us/um/redmond/groups/cue/karlson/karlson.htm, Amy K. Karlson
  7. http://innovis.cpsc.ucalgary.ca/People/SheelaghCarpendale, Sheelagh Carpendale