Bridging the communication rate gap: enhancing text input for augmentative and alternative communication (AAC)

Over 70 million people worldwide face communication difficulties, with many using augmentative and alternative communication (AAC) technology. While AAC systems help improve interaction, the communication rate gap between individuals with and without speaking difficulties remains significant, and this has led to a low sustained use of AAC systems. The study reported here combines human computer interaction (HCI) and language modelling techniques to improve the ease of use, user satisfaction, and communication rates of AAC technology in open-domain interactions. A text input interface utilising word prediction based on BERT and RoBERTa language models has been investigated with a view to improving communication rates. Three interface layouts were implemented, and it was found that a radial configuration was the most efficient. RoBERTa models fine-tuned on conversational AAC corpora led to the highest communication rates of 25.75 words per minute (WPM), with alphabetical ordering preferred over probabilistic ordering. It was also found that training on conversational corpora such as TV and Reddit outperformed training based on generic corpora such as COCA or Wikipedia. Hence, it is concluded that the limited availability of large-scale conversational AAC corpora represent a key challenge for improving communication rates and robust AAC systems.