Successfully reported this slideshow.
Your SlideShare is downloading. ×

Accessible Next Level Visualizations

Ad
Ad
Ad
Ad
Ad
Ad
Ad
Ad
Ad
Ad
Ad
Loading in …3
×

Check these out next

1 of 35 Ad

Accessible Next Level Visualizations

Download to read offline

Highcharts and Elsevier share recent research into making interactive web charts more accessible. Our usability studies focused on three areas, including stacked column charts, scatter plots, and charts with drill-down interactivity. We will share design considerations for keyboard navigation and the understandability of non-visual representations of data visualizations.

Highcharts and Elsevier share recent research into making interactive web charts more accessible. Our usability studies focused on three areas, including stacked column charts, scatter plots, and charts with drill-down interactivity. We will share design considerations for keyboard navigation and the understandability of non-visual representations of data visualizations.

Advertisement
Advertisement

More Related Content

Similar to Accessible Next Level Visualizations (20)

Advertisement

Recently uploaded (20)

Accessible Next Level Visualizations

  1. 1. Accessible Next Level Visualizations Ted Gies Øystein Moseng CSUN 2022 Friday, March 18, 2022 - 10:20 AM PST ELITE 1-3 Customized icon/illustration or, an image of Chart / Photo bit.ly/a11ynextviz
  2. 2. Sharing our research Why Accessible Data Vis? Drill Down Charts Scatter Plot UX (Text Description, AI Description, Sonification, Tactile Display) Trend Line Approach to Large Data Sets Screen Reader Demos! User Feedback from 5 Screen Reader Users Future Directions
  3. 3. Elsevier is a global information analytics business specializing in science and health
  4. 4. Elsevier + Highcharts (Since 2015)! • Started our collaboration by picking a fight with data tables • SVG accessibility, augmented with HTML • Sonification, annotations, dynamic data, interactivity • What now?
  5. 5. “As a person who is blind, how do I know if it’s safe to travel locally if I can’t see the data in a COVID cases graph?”
  6. 6. “Our results show that due to the inaccessibility of online data visualizations, screen-reader users extract information 61.48% less accurately and spend 210.96% more time interacting with online data visualizations compared to non-screen- reader users.” (Sharif et al)
  7. 7. Drill Down to look for something on a computer or website by moving from general information to more detailed information -CambridgeDictionary.org to look at or examine something in depth -Collinsdictionary.com
  8. 8. Drill Down in a Bar Chart
  9. 9. Drill Down Approach/Demo Navigated to details for China. New bar series with 5 bars. Back to Country exports button. Group: China, bar series 1 of 1 with 5 bars. Currently at bottom level. Button: Phone system devices, 223.2 billion USD Group: Country exports, bar series 1 of 1 with 4 bars. Currently at top level. Button: China, 2591 billion USD. Click to load detailed data.
  10. 10. Takeaways? • Semantics: Are the interactive bars images, buttons, or links? • Which screen reader mode? Cues for how to navigate? • How do I get back to the last View? • Which level am I at? How many levels are there? • Verbosity – “Button: China, 2591 billion USD. Click to load detailed data.” “I see, so now it gives me the components that make up that bar.”
  11. 11. Scatter Plots
  12. 12. Scatter Plots Are Used to: Demonstrate the relationship between two variables (e.g., air temp and time of day or month) (Weather – air temp lowers after 5pm in the day, goes up again after sunrise) Identification of correlational relationships and patterns Can use a mathematical model to predict weather temps in the future based upon past data
  13. 13. Challenges with Perceiving Non-Visual Scatter Plots • The term “scatter plot” • Machine readability of the graph • Information overload • Freedom of exploration • Data density • Discovery of trends, shapes and clusters
  14. 14. Our Approaches to Accessible Scatter Plots 1. Marking Up SVG data points 2. Table Version 3. Human Text Description Novel Trend Line Approach 4. Computer Generated Description 5. Sonification 6. Braille Display/Tactile Tablet Available in Highcharts today
  15. 15. Trend Line Approach An averaging algorithm to collapses several data points in a segment down into a single point. The trend line is drawn through the resultant/averaged points.
  16. 16. Scatter Demo 1 – Human Text Description We see how the winning times overall improve for males and females over time. Male marathon times start in 1897 and show a gradual linear decrease, with some variation. Female marathon times start later - in 1966 - and show a sharper non- linear decrease, which looks like a half of the letter U. After 1980 more marathons are run by both males and females. Overall, male winning times are faster. Since the late 1980s, male and female winning times have decreased at a similar rate.
  17. 17. Human Text Description Take Aways • Users overall responded positively to human created description. • Describing the shape of the data/curve was deemed useful (“half of the letter u”) • One size fits all approach might be a challenge (context: deep research or for general interest?) • May eliminate personal bias of what we think someone might want to hear
  18. 18. Scatter Demo 2 – Computer Generated Description The second data series is showing Female class, with 233 data points Female class data starts at Year 1966, where Winning time is highest on average, averaging around 03:18:53. From there Winning time goes down sharply until around Year 1978, averaging around 02:45:54. There are 23 points in this segment
  19. 19. Computer Generated Description Takeaways “The AI description was cool, but maybe too detailed and kind of vague”. Users wanted to go from Less Detailed > More Detailed. Start with a less detailed human description then go into the computer description. Limited freedom of exploration Hard to make out the shape in a linear description. Separate the computer description, don’t put it inside the actual chart data. Would be amazing if we could describe the overall shape/direction in a meaningful way.
  20. 20. Scatter Demo 3 – Sonification
  21. 21. Sonification Takeaways • Trend line concept wasn’t obvious to people who can’t see it • Good for identifying trends, but not for getting precise reads of data points • Users could discern basic trend for Y value: upward, downward and change in directions using sound • Stereo panning L/R also provided a sense for a second dimension (moving along the x axis left to right) • Some users wanted “scaffolding” to help relate sounds to a baseline set of values • Avoid screen reader announcements on top of sonification • Most users thought that being able to step through the sonification would be useful • Users had conflicting preferences
  22. 22. Scatter Demo 4 – Tactile / Braille Tablet Demo of the Peridot Tablet 48 X 64 dot resolution Featuring Dave Schleppenbach and Dr. Greg Williams
  23. 23. “using a trendline to analyze is very useful, because you're pulling out information that you would not be able to see in the scatterplot.”
  24. 24. Tactile Takeaways • Best version in terms of freedom of exploration, ability to discover trends and shapes. • Dots can blur together into “blobs”. • Users will need to pan, zoom, tuning threshold detection. (small window like a braille display) • Circles and square markers work better than other more complicated shapes. • How to disambiguate multiple data series (blinking dots). • Promise to provide a rich, co-operative learning tool between 2 people. • Unfortunately, tactile tablets are expensive $$$.
  25. 25. Summary of 7 Approaches to Accessible Scatter Plots SVG data point aria labels Table Version Sighted Person Real Time Description Human Created Description AI Description Tactile Display Sonification Pros Easy to query Table Reading Commands are quick and sophisticated Easy to query Contextually relevant description Can provide a more meaningful interpretation of data No additional effort by content creators Freedom of exploration 2-D or 3-D data representation 2-D or 3-D data representation Cons No sense for trends Bias of people predicting what they think you want to know Reliance on another human Not always contextually relevant Requires advanced math and algorithms No one size fits all approach Expensive Requires some training (e.g., adjusting threshold detection) Difficult to layer different data in same graph Need scaffolding context, training Identify Trends? No No Yes Yes Emerging Yes Yes Freedom of Exploration? Limited Yes No No Limited Yes Limited
  26. 26. Future Directions • Tactile display with announcement of data on PC • Accessibility Menu • Sonification tools • role=“application”? • Breadcrumbs for drill down • Highcharts authoring accessibility checker
  27. 27. Thank You Ted Gies Øystein Moseng www.linkedin.com/in/tedgies www.linkedin.com/in/oysteinmoseng
  28. 28. Thank you to our users/collaborators Dr. John Gardner (ViewPlus) Dr. Nicholas Giudice (U of Maine) Lucy Greco (UC Berkeley) Hadi Rangin (U of Washington) David Schleppenbach (Tactile Solutions) Ryan Shugart (Microsoft) Dr. Greg Williams (Independence Science)
  29. 29. References Highcharts accessibility demos Highcharts accessibility portal Highcharts accessibility module Sonification API Highcharts Features History YouTube: Highsoft and Elsevier Talk About Accessibility 2020 Understanding Screen-Reader Users’ Experiences with Online Data Visualizations (Sharif, et al) Contact Us Ted Gies <ted.gies@elsevier.com> Øystein Moseng <oystein@highsoft.com>
  30. 30. Previous CSUN Sessions VPATs for Business or Measure Accessible SVG charts using ARIA Elsevier: Article of the future collaboration Accessibility for Large Publishers: Challenges, Choices, Change Alt Text – A Process of Discovery Don’t Play Me – 2 Games in Web Accessibility Elsevier Company Accessibility Policy
  31. 31. EXTRA SLIDES
  32. 32. Our 6 Approaches to Accessible Scatter Plots SVG data point aria labels Table Version Human Created Description AI Description Sonification Tactile Display Identify Trends? No No Yes Emerging Yes Yes Freedom of Exploration? Limited Yes No Limited Limited Yes Novel Trend Line Approach
  33. 33. A. I’ve never experienced an accessible visualization B. A Table C. Have a sighted person explain it to me real time D. Human created text description E. AI generated description F. Tactile Display G. Sonification Poll: What is the best way to experience accessible visualizations on the Web?

×