Efficient Artificial Intelligence-Teaching Assistant Based on ChatGPT

Abstract

Due to the real problem of “insufficient basic knowledge in interdisciplinary fields” that students may face when studying interdisciplinary courses and learning interdisciplinary knowledge, this study proposes an “Artificial Intelligence (AI)-Teaching Assistant (TA) App” system. The proposed system combines a Question-and-Answer (QA) system, an expansion mechanism, a cache mechanism, and an error correction mechanism to assist students in their learning process by AI-based QA system. Students can ask “AI-TA App” questions through voice and the App can quickly provide corresponding answers to help them learn. For providing an efficient “AI-TA App,” this study proposes a cache mechanism. A neural network-based QA system is built in the “AI-TA App” server and database to enable the QA system to respond to students' questions quickly; experimental results show that the average response time using the cache mechanism is 0.001 seconds, while the average response time using ChatGPT is 15.652 seconds, indicating that the proposed cache mechanism in this study can effectively improve system efficiency. Furthermore, this study also proposes an expansion mechanism. By calling the ChatGPT APIs, the expansion mechanism expands the question bank and improves the accuracy and robustness of the QA system; experimental results show that the accuracy without using the expansion mechanism is 25.45%, while the accuracy can be improved to 99.09% by using the expansion mechanism.

Ähnliche Arbeiten