SocializeChat
A GPT-based AAC Tool for Social Communication
Yunkai Xu, Yuyang Fang, Zhaoqu Jiang, Zhuyu Teng
Teamwork role: researcher, developer, user tester.
People experiencing speech impairment and sometimes even motor impairment often encounter significant communication challenges, resulting in the reliance on Augmentative and Alternative Communication (AAC) tools. Unlike traditional AACs which provide basic words of physiological needs, the Large Language Model provides possibilities for diverse and engaging social conversations. However, existing AAC tools and applications can hardly handle the social context considering the limited input from AAC users and the varied relations among conversation partners. We developed SocializeChat, a GPT-based AAC tool, aiming to help their social communication. SocializeChat is rooted in a field study involving users with speech impairment. It employed two influential factors in social communication - topic preference and closeness to generate sentence suggestions for AAC users. A user test demonstrates its strength in supporting social communication, displaying its potential to foster an inclusive social life for the users.
