A Major Recommendation System for University Admissions Based on Knowledge Graphs and Large Language Models

Most college entrance examination candidates in China face significant difficulties and confusion when filling out their university and major preferences. Existing systems that provide access to historical admission data can offer some assistance in decision-making. However, these systems typically require users to input information in a rigid, structured format, making it difficult to support natural and personalized interactions with students. On the other hand, large language models (LLMs) possess strong capabilities in natural language understanding and generation, enabling more user-friendly dialogue interfaces. Yet, due to the well-known issue of hallucination in LLMs, the reliability and factual correctness of their outputs cannot be fully guaranteed. To address these challenges, this paper constructs a structured knowledge graph of university-major admission data based on historical college entrance examination scores. This knowledge graph serves as a trustworthy information base to constrain and enhance the recommendation process. Building on this foundation, we propose a university and major recommendation framework that integrates extraction-enhanced techniques and large language models. By leveraging the knowledge graph to guide the generation process, our system improves both the accuracy and relevance of the recommendations. Experimental results demonstrate that the proposed method significantly enhances the precision of university and major recommendations, indicating its strong potential for real-world application and further development.