Enhancing Non-Visual Presentation Accessibility with Real-Time Feedback Tool (Full Transcript)

Speaker 1: Hi everyone, I'm Yihao. Today I will share our project, Say It All, feedback for improving non-visual presentation accessibility. This work was collabed with Jun and advised by Jeff and Amy. Presenters often do not describe the content on their slides. Here's an example.

Speaker 2: Anyway, so we did the study and, you know, basically the audio captures are a lot harder. Not surprising.

Speaker 1: So the presenter didn't describe his slides, which include a graph and the study information like the number of participants, which in this case was 162. When presenters do not describe the content on their slides, people who are blind and visually impaired will miss out on the information in the presentation, making it inaccessible. But how often do speakers fail to fully describe their slides in practice? To find out, we analyzed 90 presentation videos in a while. The video clips contain 269 slides and 1650 total slide elements, half of which were text and the other half are media. Two people coded the speaker's verbal coverage of each slide element as non, little, half, most, and complete. Overall, presenters didn't fully describe their slides, leaving out key information for about 72% of slide elements. When presenters are explicitly trying to create an accessible presentation, they can use guidelines like those from W3C or Seek Access. This suggests that speakers should describe all pertinent visual information on their slides and use minimal visuals. However, it can still be challenging for presenters to remember to describe their slides. So the goal of our work is to help presenters make their presentations accessible. And towards this goal, we present PresentationOutline, a tool that gives automatic feedback to help presenters describe their slides. PresentationOutline provides real-time feedback and post-presentation feedback. The real-time feedback interface augments the existing slides' presenter view to give presenters feedback on what they have and haven't described. In this video, the interface highlights the slide text, creates a colorful circle brush, and an image depicting the circle brush, as the speaker described them.

Speaker 3: First we'll create a colorful circle brush. And you'll see this squiggly line pictured on the right with these rainbow circles sort of getting wider and smaller as it goes.

Speaker 1: Users can also turn on or off the feedback display. On the other hand, our post-presentation feedback interface augments the Google Slide Editor to let users review their results. Users can quickly identify what slides they did and did not describe well by glancing at the slide overview. It shows poorly described slides in red. When editing each individual slide, PresentationOutline highlights the elements they were described during the presentation in green. A results panel shows the slide's coverage percentage, the transcript recorded during the talk, and a specific suggestion for how to make the slide more accessible. The interface updates the coverage percentage as the presenters edit the slides. Or edits the transcripts to add the descriptions. We invited 60 people to present and review their own slides with PresentationOutline. Users presented half of the slides with real-time feedback and half without. After presentation, they reviewed their slides with and without post-presentation feedback. PresentationOutline's real-time feedback helped people describe significantly more of the text image than the default interface. For text, people covered 57% of text with and 46% of text without our system. People achieved an image coverage score of 3.5 out of 5 with and 3.1 out of 5 without our system. In addition, the post-presentation feedback helped people identify significantly more accessibility changes to make in the future than they did without our feedback. People identified 2.3 changes without our system and 0.7 without. In the future, we plan to provide people more granular feedback on how to describe their image and diagrams. Also, our current interface is not accessible for blind and low-vision presenters because it relies on visual for feedback. We plan to make a screen reader accessible tool in the future. Finally, we plan to deploy the tool as a Google Chrome extension. With that, I would like to end my talk. For more information, you can find it at setitall.github.io. Thank you.