Design and Implementation of Online Testing and Evaluation System Based on Deep Learning and Knowledge Graph

In order to solve the problems of inaccurate tracking of students' learning progress and difficulty in quantifying learning outcomes, Transformer neural network models and knowledge graph technology are introduced to construct an intelligent system that can more accurately evaluate students' learning outcomes. The study first introduces the Transformer neural network model to evaluate and predict students' learning achievements through their historical learning data, then uses knowledge graph technology to construct domain knowledge graphs and learner portraits, which are respectively used to store subject knowledge systems and track students' learning progress. Finally, the system achieves intelligent question answering and evaluation through three core modules: question parsing, similarity matching, and answer generation. The system adopts natural language processing technology to process student input, uses Transformer neural network model for problem intention recognition, and combines students' abilities and knowledge graph to dynamically generate evaluation content and personalized suggestions. The data shows that the system can significantly improve the tracking accuracy of students' learning progress and the quantitative evaluation of learning effectiveness. The accuracy of learning progress tracking has increased from 70% to 91%, and the system's analysis time for students' learning outcomes has been reduced from 5 hours to 3 hours. These results validate the effectiveness of combining deep learning and knowledge graph technology, significantly improving the efficiency of educational resource utilization and the responsiveness of teaching activities.