SymCA Framework Uses LLM Symbolic Decision Processes for Column Annotation

SymCA improves column type and property annotation by materializing the task as a global-to-local symbolic decision process. It utilizes LLM-driven skeleton induction and substrate evolution to capture label semantics and improve interpretability over direct neural mapping.