accessibility and assistive technology
While applying to the PhD program I was interested in accessibility and assistive technology and how different technologies were created. Speaking with Dr. James Clawson and hearing his previous interest in text entry made me curious as to what methodologies were used for conducting text entry experiments with people who identify as low-vision or blind. I found that many researchers would read phrases from text entry sets common in the literature out loud to participants, but this made me think that a single person reading phrases designed to be read visually might not mean they are easy to understand clearly read out. So working with Dr. Jofish Kaye and Dr. James Clawson, we took a set of 500 phrases and used four different text to speech systems to generate audio clips. We then had 392 participants from Amazon’s Mechanical Turk transcribe the phrases and we reported on their errors. Of the 500 phrases, only 96 were found to not have any comprehension errors when read out loud across all the text to speech systems. Taking those remaining 96 phrases, we recruited 80 participants who identified as low-vision or blind and asked them to transcribe the audio clips. From the 96 phrases only 4 phrases were found to have any comprehension errors, leaving 92 aurally distinct phrases that maintained across both experiments and could be used in future text entry studies.
Identifying an Aurally Distinct Phrase Set for Text Entry Techniques
Jacob Abbott, Jofish Kaye, and James Clawson. In Proceedings of the 2022 CHI Conference on Human Factors in Computing Systems, ACM, April 2022. (Honorable Mention for Best Paper Award) DOI: https://doi.org/10.1145/3491102.3501897
All photography provided by Jacob E. Abbott