causal-graph

This paper proposes a hybrid framework combining LLM summarization, a linguistically grounded Expert Index, and STAC classification to generate causal graphs from narrative texts, outperforming GPT-4o and Claude 3.5 on benchmark stories.