Many individuals can benefit from augmentative and alternative communication (AAC) while hospitalized, including regular AAC users and patients with temporary communication impairments (e.g. due to intubation). However, hospitals may not have access to the wide variety of AAC solutions required to serve the entire population of individuals with severe speech and physical impairments (SSPI). Moreover, clinics and smaller hospitals may not have AAC-experienced speech language pathologists on call. Rather than prescribe a one size fits all solution, we have constructed a system that learns and adapts to the user’s capabilities. When access methods are less dependable, we automatically adjust the query pacing so that users in need are given more opportunities to express themselves. In other words, when a user’s physiology precludes them from accurate switch (sip-and-puff, button switch, among others) input, we aggregate the more uncertain inputs until we reach sufficiently high confidence in a letter to move forward in their typing task. By doing so, we support accurate letter prediction for users with less-than-trustworthy access methods. Our system builds on the strength of DASHER but adds stepwise user input (no continuous, fast-paced interface is needed) and explicit error modelling to adapt to each user’s capabilities. Shuffle speller has the potential to enable individuals with SSPI to communicate with their loved ones, relay medically relevant messages, health-care decisions, and set environmental control preferences, which are vitally important in hospital settings.
Fernando Quivira and Matt Higger won the 2017 RERC on AAC Student Research and Development Competition