A question-and-answer system based on artificial intelligence

This study is the research outcome of the project "Exploration of Student-Centered Teaching Reform and AI Technology Application in Dynamic Website Design Courses" led by the author at ZhuJiang College, South China Agricultural University. This study investigates the development and application of an intelligent question-and-answer (Q&A) system tailored for computer science course instruction, using the BERT model as its core framework. Traditional teaching methods face limitations in providing timely, personalized responses to students' inquiries, especially in large classroom settings. To address this, we propose a Q&A system based on BERT's bidirectional training capabilities, enabling deeper contextual understanding and precise response generation. The research introduces specific advancements in system structure, model selection, dataset creation, and training processes to address the unique needs of computer science education. The methodology encompasses dataset construction, using a collection of 2,000 carefully curated question-answer pairs spanning essential topics like operating systems, programming languages, and databases. We refined the BERT model with this dataset through fine-tuning, employing dynamic masking and the Adam optimizer with a decay learning rate to achieve optimal generalization. Furthermore, we developed a user feedback mechanism, where responses are evaluated for accuracy, enabling continuous model refinement and adaptation to student feedback. This study designed an MVC-based system integrating Java, TorchServe, and MySQL. Experiments show that it outperforms traditional methods in performance and efficiency, achieving an F1 score of 91.1% with a response time of just 0.8 seconds. Error analysis indicates room for improvement in handling complex terminology and contextual understanding. This work's innovations contribute a robust, scalable tool for enhancing computer science education, offering educators a powerful solution to improve teaching interactivity, efficiency, and personalized learning support. The insights gained from this study support broader applications of AI-driven Q&A systems in academic settings, paving the way for future developments in automated educational assistance.