In healthcare management, the manual assessment of documents to verify compliance with legal regulations is labor-intensive, error-prone, and time-consuming. These documents often consist of multiple paragraphs, and the accurate annotation of these paragraphs is essential for ensuring alignment with specific legal requirements and facilitating efficient regulatory oversight. To address these challenges, this research proposes a novel approach for automating the annotation of paragraphs within care support documents using Natural Language Processing (NLP) models and a knowledge graph (KG) to compute semantic similarity between legal articles and document paragraphs, enabling efficient annotation for compliance. Experiments on real-world healthcare documents show the method significantly improves annotation accuracy, reduces compliance verification time, and enhances similarity confidence when integrating a KG. These results highlight the potential of combining NLP and KGs to automate compliance assessment in healthcare.