The existing automatic chalkboard writing evaluation system developed by our research team has already achieved automatic comparison, scoring, and assessment. In response to the current system’s limited interactivity, we introduce a dialogue-based framework for normative chalkboard writing, supported by large language models. This system aims to provide personalized guidance and concrete practice suggestions. In order to improve the understanding and task performance of large language models in normative chalkboard writing, we leverage retrieval-augmented generation (RAG) to integrate relevant external knowledge. Moreover, we refine the traditional RAG pipeline by incorporating advanced strategies such as hybrid retrieval, re-ranking, and query rewriting. Finally, we conduct a comprehensive evaluation of the proposed approach using the open-source RAGAS framework. Experimental results show that our method significantly outperforms baseline models.