Design of Artificial Intelligence Response System for Virtual Teaching and Research Room based on Improved TDNN Model

Based on the improved Time Delay Neural Networks (TDNN) model, this study aims to address the issues of poor user experience and low intelligence in the current intelligent answering system for virtual classrooms, and to conduct research on teacher student interaction for teaching research. This system can identify the semantics and logic of massive teaching resources, teacher-student interaction records, course requirement texts, etc., extract the relationships between entities (courses, questions, requirements, etc.), and achieve semantic retrieval, intelligent question answering, and visual teaching knowledge retrieval in the field of teaching research. This system integrates various artificial intelligence technologies, forming a hybrid mode that allows virtual assistants to accurately understand students and their needs. The knowledge map connects teaching research with knowledge in teaching research, and represents its internal logic, making the connections between knowledge more clear. The knowledge map opens up its own capabilities and outputs entity relationships to the knowledge base and intelligent virtual assistants, making knowledge retrieval and answering faster. The experimental results demonstrate that the improved TDNN model exhibits continuously improving accuracy during the testing process. As the training progresses, the accuracy of the model gradually improves, increasing from the initial 92.13% to the final 97.05%.