The exponential growth of table data across various domains has underscored the critical need for accurate annotation of table metadata, particularly column types and inter-column relationships. Addressing this challenge, we introduce KETAM, a Knowledge-Enhanced Pre-trained Language Model that integrates domain-specific knowledge with pre-trained language models to enhance the precision and effectiveness of table metadata annotation. KETAM employs a novel methodology that fortifies table data with domain insights, leveraging these to annotate column types and inter-column relationships with unprecedented accuracy. Through extensive testing on the WikiTable and VizNet datasets, KETAM demonstrates superior performance, achieving a Macro F1 score of 86.62% and a Micro F1 score of 94.37% for column type annotation on the WikiTable dataset, and a Macro F1 score of 84.68% and a Micro F1 score of 93.26% on the VizNet dataset, significantly outperforming existing models like TURL and Sato. Similarly, for column relationship annotation on the WikiTable dataset, KETAM obtains a Macro F1 score of 84.91% and a Micro F1 score of 93.70%, surpassing TURL by a notable margin. These robust empirical results validate KETAM’s superior performance and reliability in annotating table metadata, showcasing its potential to revolutionize data analysis through advanced table understanding.