The CTU-Helper system is an automated question-answering system designed to assist students and prospective applicants at Can Tho University (CTU) in quickly and efficiently accessing information regarding academic reg- ulations and admissions. This system utilizes Retrieval-Augmented Generation (RAG) combined with Bi-encoder and Cross-encoder models for natural language processing. The Bi-encoder allows the system to compare user questions with questions stored in the database, thereby identifying corresponding answers. The Cross-encoder is used to assess the relevance between user questions and potential answers, ensuring the accuracy of returned results. Additionally, CTU-Helper is equipped with Semantic Router technology, which facilitates semantic classifica- tion for queries. This technology enables the system to identify the main topic of the question and forward the query to the appropriate processing model, enhanc- ing information retrieval efficiency. Evaluation results indicate that CTU-Helper achieves high performance with a Recall of 0.8683 in the information retrieval process. This demonstrates the system’s ability to effectively search and return accurate information to users.

With its outstanding advantages, CTU-Helper promises to be a valuable sup- port tool for students and applicants at CTU, facilitating easier and more conve- nient access to information. This system also opens up new research directions in applying advanced natural language processing techniques to build automated question-answering systems in the educational field.