Storyline Trees: Hierarchical Representations for Long-Form Narratives

2026-06-18

AI Summary

This work addresses the challenge that long-form narrative texts often lack explicit structure, hindering long-context models from effectively reasoning about events, characters, and plot progression. To overcome this limitation, the authors propose organizing narratives into a hierarchical “storyline tree” with scenes as the fundamental unit, integrating top-down and bottom-up strategies to capture multi-granular structure—from global themes to fine-grained events. An adaptive retrieval mechanism is further introduced to enhance question-answering performance without increasing retrieval overhead. Evaluated on three long-context narrative question-answering benchmarks, the proposed approach significantly outperforms strong baselines, including post-trained long-context models and agent-based chunking methods, thereby demonstrating the efficacy of scene-based units and storyline tree representations for narrative understanding.

Abstract

Long-form narratives are challenging for long-context models because their structure is implicit: events, characters, and plotlines interact across hundreds of pages without the explicit cues that guide navigation in structured documents. We address this by constructing storyline trees, hierarchical representations that organize narratives from global themes and major plotlines to fine-grained events. We first segment chapters into contiguous narrative segments, or scenes, and use them as the basic units for tree construction. We then infer storyline trees through complementary top-down and bottom-up procedures that derive, refine, cluster, and summarize storylines at multiple levels of abstraction. We showcase the utility of this representation for question answering: storyline trees enable adaptive retrieval, allowing models to iteratively inspect high-level narrative structure and retrieve scene-level evidence on demand. Experiments on three long-context narrative QA benchmarks show that adaptive retrieval outperforms strong baselines, including post-trained long-context models and agentic chunk-based methods. Ablations confirm that scenes are more effective basic units than chapters or generic segmentation, and that gains persist under matched retrieval budgets

Problem

• long-form narratives
• implicit structure
• storyline representation
• narrative understanding
• long-context models

Innovation

• storyline trees
• hierarchical narrative representation
• adaptive retrieval
• scene segmentation
• long-form narrative QA