textual-data

This paper proposes LMT, a Bayesian causal discovery framework that combines LLM-extracted semantic signals from textual alarm records with timestamp-based statistical evidence to infer causal graphs in manufacturing systems.