Speaker: Eric Wright
Twitter: @ewaccess
Note: During this notes AT refers to: Assistive Technology
Dragon -> AT to replace the keyboard for input. [Insert link here]
“Most speech databases are trained with an overrepresentation of certain demographics which results in a bias towards male and white and middle class. Accents and dialecs tend to be underrepresented in training datasets. Many machines struggle with understanding female voices or voices of elderly people” - Kelly Davis, Head of Machine Learning , Mozilla
Mozila Common Voice Project isinviting people to dictate with their own voice to increase / fix the training dataset
https://voice.mozilla.org/en
Now image if we have a close button with an X as text
<button aria-label="Close">X</button>
Screen reader announces this as “Close” (thanks to the aria label).
Using Dragon, the user can say:
In Dragon, aria-label adds to the content.
Speech recognition requires a role
<a href="https://en.wikipeidia.org/wiki/Yes_(band)">
<img src="https://upload/wikimedia.org/wikipedia/commons/5/51/Green-checkmark.svg title="Correct!" alt="Yes!" aria-label="System requirements met" aria-labelledby="nope/>
</a>
<span class="hidden" id="nope">ridiculous</span>
Screen reader announces: Ridiculous. [aria-labelledby] Link graphic [roles]
As Dragon user I can say:
<label>Name: <input type="text"/></label>
This won’t work on Dragon, even though it’s perfect HTML.
<button aria-label="Close">X</button>
What I see: A close button
What I say: press ESC button
How would you design for a keyboard user who is not pressing anything?