Charts and graphs. Whether its Covid statistics, budgeting info or any other numerical or statistical data, increasingly, charts and graphs are being used to present information in a consumer-friendly manner. “research shows that people don’t want to read text—all they want is a quick visualization in an attractive chart or graph,” says Ather Sharif, a University of Washington doctoral student in the Paul G. Allen School of Computer Science & Engineering. A heat map of voter turnout is ever so much easier to digest than a 500-cell table containing the same data set. “If I’m looking at that heat map I can pull out whatever information I am interested in,” says Sharif, “Maybe it’s my own state, or maybe it’s the overall trend or the maximum values."
A text-based table may provide the screen reader or braille user the raw data, but it does a poor job of presenting the actual information. Sharif makes a distinction between equality and equity. “If you provide somebody with a table you are making it equal as far as the information. But imagine a table containing 10,000 rows or more.”
When web developers accompany charts and graphs with data tables they’ve only made visible to screen readers it’s up to the user to mentally calculate the minimum, maximum, mean and medium—information that’s fairly obvious to the sighted user. And if it’s a bell curve listing standard deviations unless you download that table and run it through a spread sheet you are basically out of luck.
“Developers often fail to consider the cognitive load this puts on the user,” says Sharif. “They offer equal data, but there’s little equity in the one-dimensional text experience.”
“There is a start and an end of a sentence and everything else comes in between,” adds Jacob O. Wobbrock, UW professor in the Information School. “But as soon as you move things into two dimensional spaces, such as visualizations, there’s no clear start and finish. It’s just not structured in the same way, which means there’s no obvious entry point or sequencing for screen readers.” Says Sharif, “If two people look at the same chart they are usually having two separate experiences. One individual might hone in on the average, or mean, and the other might spot an unusual distribution or a particular outlier. So how can we justify a single alt text label for that chart?” he asks. “Unfortunately, screen reader and braille users have been told that tabular data is their best option. Few have actually stopped to ask, ‘Is there something that could be better than best?’” Sharif and his team members think they've come up with a solution. It’s called VoxLens, and here’s how it works.
Another current limitation is that VoxLens only works with visualizations in a single dimension. For example, the program can digest and interpret a daily stock chart, the highs and lows, overall averages, etc., but cannot currently work with, say, a price line compared to a moving average of daily closing prices. Currently VoxLens supports the following commands: maximum, minimum, average, mean, median, mode, standard deviation, and total. “More importantly, you can extract individual data points,” says Sharif. “For example, in the sample graphs you can say ‘What is the value for California?’ You can also ask for more than one data point. For example, ‘what is the value for California and Illinois?’ Need help? Say "instructions," or "directions," or "help" to hear the instructions. You can also say "commands" to hear all the list of supported commands.
For now, VoxLens is only available on web pages—e-textbooks and PDFs are not supported. “We are hoping to add these capabilities and more as we move forward,” says Sharif.
Are you in contact with a web developer who might be willing to enhance their visualizations with VoxLens? Consider clicking the share button and sending a copy their way.
Where to learn more
- The VoxLens Open-source GitHub Repo
- VoxLens examples
This article is made possible in part by generous funding from the James H. and Alice Teubert Charitable Trust, Huntington, West Virginia.