A Multimodal Alerting System for Online Class Quality Assurance
Project Overview
The document explores the integration of generative AI in education, particularly through the development of a multimodal alerting system designed for online 1-on-1 classes in China to enhance quality assurance. This innovative system incorporates a banned word detector and a class quality predictor, leveraging multimodal data from teaching sessions to monitor instructor behavior and assess the overall quality of courses. It responds to the increasing reliance on inexperienced part-time instructors by providing critical oversight and ensuring that educational standards are upheld. The findings highlight the potential of such AI applications to balance teaching resources effectively while maintaining a high quality of education, addressing challenges that arise in the rapidly evolving landscape of online learning. Through these advancements, the document underscores the transformative impact of generative AI in fostering better educational outcomes and supporting instructors in delivering effective teaching experiences.
Key Applications
Multimodal Alerting System
Context: Online 1-on-1 tutoring for K-12 students in China
Implementation: The system uses audio and video data from classes to analyze instructor behavior through a banned word detector and evaluate course quality with a quality predictor. It outputs alerts based on detected issues.
Outcomes: Achieved 74.3% alerting accuracy in identifying low-quality classes and misbehavior.
Challenges: Part-time instructors may lack experience, leading to potential quality issues that traditional rating systems do not address effectively.
Implementation Barriers
Quality Assurance
Ensuring the quality of classes taught by inexperienced part-time instructors.
Proposed Solutions: Implement a multimodal alerting system that detects banned words and predicts class quality using linguistic and prosodic features.
Project Team
Jiahao Chen
Researcher
Hang Li
Researcher
Wenxin Wang
Researcher
Wenbiao Ding
Researcher
Gale Yan Huang
Researcher
Zitao Liu
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Jiahao Chen, Hang Li, Wenxin Wang, Wenbiao Ding, Gale Yan Huang, Zitao Liu
Source Publication: View Original PaperLink opens in a new window
Project Contact: Dr. Jianhua Yang
LLM Model Version: gpt-4o-mini-2024-07-18
Analysis Provider: Openai