Inside CORE-KG: Evaluating Structured Prompting and Coreference Resolution for Knowledge Graphs