DigitCHAT: Enabling AAC Input at Conversational Speed
Karl Wiegand, Rupal Patel · 2014 · Proceedings of the 16th International ACM SIGACCESS Conference on Computers & Accessibility (ASSETS 2014) · doi:10.1145/2661334.2661422
Summary
This paper presents DigitCHAT, a prototype augmentative and alternative communication (AAC) system designed for literate users with mild upper limb motor impairments who need fast, expressive face-to-face communication on mobile phones. The authors identify a gap in existing AAC systems: most mobile AAC apps use on-screen keyboards with small, adjacent buttons that are difficult for people with motor impairments or vision limitations to use accurately. Additionally, conventional AAC systems tend to focus on pre-stored utterances rather than real-time composition, and require listeners to wait while the user composes a full message — creating uncomfortable silences and pressuring users toward telegraphic speech. DigitCHAT addresses these issues with a telephone number pad layout featuring large, visually separated buttons. The familiar 3x3 grid reduces learning time compared to full QWERTY layouts while accepting higher lexical ambiguity as a trade-off. The system supports two input modes: mixed mode (discrete taps or non-contiguous path segments) and continuous mode (drawing a single line through all desired buttons, similar to Swype-style gesture typing). A dictionary-based implementation using unigram statistics from the Crowdsourced AAC-Like Corpus maps physical paths across the number pad to candidate words. Critically, DigitCHAT speaks each word aloud immediately upon completion rather than waiting for a full utterance, enabling more natural conversational turn-taking.
Key findings
The system resolves textonyms — words sharing the same button sequence, like "bat" and "cat" — through two mechanisms: users can scribble over the last button to cycle through alternatives, or the system learns preferred textonyms over time. The word-at-a-time speech output model is a significant departure from conventional AAC systems that require complete utterance composition before output, directly addressing the social dynamics of conversation by reducing uncomfortable silences. The app was released on the Google Play Store and underwent two design iterations based on user feedback from the target population, including narrative feedback from three users with chronic neuromotor disorders. A key user-requested feature was cancellation capability to prevent undesired words from being spoken. The authors also created configurable options such as movement thresholds for textonym rotation and were preparing for a formal clinical study.
Relevance
DigitCHAT highlights important but often overlooked social dimensions of AAC design — stigma around carrying conspicuous devices, the pressure of conversational silences, and the need for discreet, phone-based solutions. For accessibility practitioners, the paper demonstrates how familiar interaction paradigms (telephone keypads, gesture-based input) can be adapted for AAC use on mainstream consumer devices. The continuous motion input approach accommodates users who can maintain contact with a screen but struggle with precise tap targeting. The word-by-word speech output model offers a compelling alternative to utterance-level AAC systems for face-to-face conversation. Future directions include skip-gram language models, phrasal prediction, and full vocabulary customization.
Tags: augmentative and alternative communication · AAC · text entry · mobile accessibility · motor impairments · continuous motion input · speech impairments · conversational interaction