Date of Award
7-17-2025
Document Type
Thesis
Publisher
Santa Clara : Santa Clara University, 2025
Department
Electrical and Computer Engineering
First Advisor
Hoeseok Yang
Abstract
The number of people diagnosed with Autism Spectrum Disorder is growing annually at a rapid pace. Autistic people are known to have multiple problems with communication, requiring assistance and coaching which is already limited in availability due to the amount of training required. At the same time, there has been a recent exponential rise in usage of LLMs like ChatGPT which is widely available. This project aims to prototype a solution using ChatGPT to provide communication assistance to those diagnosed with autism. The project would record voice input through a microphone and transcribe it into text. The text would be inserted into an annotated prompt and sent to ChatGPT, which would then generate a response as text and displayed to guide conversation in real time. The prototype would also manage pace, detect silence, and re-generate advice when needed. These features would facilitate a conversation between an autistic person and non-autistic person. Analysis shows that selecting the appropriate models helps improve usability by reducing latency, and fine-tuning prompts helps improve advice quality. Further refinements in hardware, software, and cloud components can be explored to enhance performance for communication assistance.
Recommended Citation
Li, Justin Y., "Communication Assistance Using A Large Language Model" (2025). Electrical and Computer Engineering Senior Theses. 103.
https://scholarcommons.scu.edu/elec_senior/103
