LMT: A Bayesian Framework for Causal Discovery from Textual Alarm Records in Manufacturing Systems

In three linesLMT is a Bayesian causal discovery framework for manufacturing alarm logs that jointly leverages LLM-extracted semantic signals from event descriptions and temporal evidence via Poisson-process likelihood. It refines LLM-informed priors with timestamp-based statistical evidence to produce interpretable, data-supported causal graphs. Simulation studies demonstrate effectiveness in small-sample scenarios.

Summary generated by Claude — human-verified