CREFT: Sequential Multi-Agent LLM for Character Relation Extraction

2025-05-30

AI Summary

Existing methods struggle to effectively parse fine-grained, dynamically evolving character relationships in long-form narratives, hindering narrative analysis and script evaluation. To address this, we propose a multi-stage collaborative LLM agent framework that unifies character graph construction, relationship refinement, role identification, and group partitioning. Our key contributions are: (1) a staged, specialized LLM agent coordination mechanism enabling progressive reasoning; and (2) the integration of knowledge distillation with iterative graph-structure optimization to alleviate long-range dependency modeling bottlenecks. Evaluated on a Korean drama script dataset, our framework achieves significantly higher accuracy and completeness in relationship extraction compared to single-agent baselines. Moreover, it enables interpretable character network visualization and facilitates efficient, scalable script assessment—demonstrating both analytical rigor and practical utility for narrative understanding.

Technology Category

• Multiagent Systems: Coordination and Collaboration
• Machine Learning: Large Multimodal Models (LMMs)
• Cognitive Modeling & Cognitive Systems: Agent Architectures

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Understanding complex character relations is crucial for narrative analysis and efficient script evaluation, yet existing extraction methods often fail to handle long-form narratives with nuanced interactions. To address this challenge, we present CREFT, a novel sequential framework leveraging specialized Large Language Model (LLM) agents. First, CREFT builds a base character graph through knowledge distillation, then iteratively refines character composition, relation extraction, role identification, and group assignments. Experiments on a curated Korean drama dataset demonstrate that CREFT significantly outperforms single-agent LLM baselines in both accuracy and completeness. By systematically visualizing character networks, CREFT streamlines narrative comprehension and accelerates script review -- offering substantial benefits to the entertainment, publishing, and educational sectors.

Problem

• Extracting complex character relations in long narratives
• Improving accuracy of multi-agent LLM relation extraction
• Streamlining narrative analysis for entertainment industries

Innovation

• Sequential multi-agent LLM framework
• Knowledge distillation for base graph
• Iterative refinement of character relations