Designing a Context Aware AAC Solution
Conor McKillop · 2018 · Proceedings of the 20th International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS '18) · doi:10.1145/3234695.3240990
Summary
This student research competition paper explores the design of a context-aware AAC keyboard prototype that uses mobile device sensors — particularly GPS and Wi-Fi positioning — to suggest location-relevant phrases and categories for people with speech, language, and communication needs. Most AAC devices achieve communication rates below 10 words per minute, far less than unimpaired speech, and their word prediction relies mainly on past usage without considering the user's current context. The author identifies four categories of communication context — location, identity, time, and activity — and notes that very few existing AAC devices utilize context-aware functionality. The prototype, developed at the University of Dundee, displays conversational options relevant to the user's current location: at a medical centre, healthcare categories like 'Medication,' 'Doctor,' and 'Disabled Access' are prioritized; at a leisure centre, phrases like 'Where are the changing facilities?' would surface. The design followed an iterative, user-centered approach involving participants from the University of Dundee's Straight Talking Group (STG), progressing from paper prototypes through requirements gathering to a high-fidelity interactive Axure prototype.
Key findings
A focus group evaluation with seven participants (four AAC users with cerebral palsy aged 38-61, plus three support workers) yielded unanimously positive feedback. Participants reported frustrations with current devices: slow word prediction, prediction software learning incorrect words (including misspellings), inability to delete whole sentences, and voice synthesis mispronouncing names of people and places. In contrast, all agreed the context-aware prototype could improve their communication rate and ease. Specific use cases resonated strongly: participants valued that location-aware phrases at home could help them communicate needs to new care staff without retyping the same sentences repeatedly, and that finance-related phrases could be automatically surfaced at a bank. One participant noted the context-aware prediction would help them select words they normally struggle to spell. While no negative aspects of the prototype were raised, the author acknowledges that group-based evaluation may have introduced social desirability bias.
Relevance
This paper tackles a practical gap in AAC technology: despite the proliferation of context-aware computing in mainstream mobile applications (location-based suggestions, time-aware notifications), AAC devices have largely not adopted these capabilities. The concept of using location to pre-populate relevant vocabulary categories and phrases is straightforward yet potentially high-impact — reducing the navigation and typing burden for common, predictable communication scenarios like healthcare visits, banking, or interacting with care staff at home. For practitioners, the user feedback reveals that AAC users' frustrations extend beyond communication rate to include prediction quality (learning wrong words), text editing limitations (character-by-character deletion), and speech synthesis accuracy (mispronouncing proper nouns) — all areas where modern NLP and TTS could significantly improve the experience. The participatory design approach with the Straight Talking Group demonstrates the value of involving AAC users throughout the design process rather than designing for them.
Tags: augmentative and alternative communication · context awareness · word prediction · cerebral palsy · mobile accessibility · speech generating devices · communication · participatory design · user-centered design