Critics eye uses graph reasoning to find motifs in long stories
Critic's Eye: Review-Grounded Graph Reasoning for Long-Form Motif Discovery
Artificial Intelligence
Summary
Understanding stories deeply means noticing themes that show up again and again, not just the order of events. The authors study how repeated ideas like 'blood' in Macbeth carry meaning across the whole story, which isn't easy to spot with simple models. They created Critic's Eye, a method that looks at the story as a big graph of connections to find these motifs more accurately. This approach works much better than big, general AI models that just process sequences of text in order.
What this means in practice
- •For literary analysis tools: Enhance automated literary software to identify themes and motifs across entire works using graph-based reasoning.
- •For digital humanities projects: Incorporate structural narrative graphs to deepen analysis of symbolic patterns in classical and modern texts.
Authors
Xinyu Sun, Zhangyue Yin, Xinyi Ju, Xin Wang
Abstract
Deep narrative understanding requires distinguishing what happens from how it is presented, moving beyond the chronological fabula to decode the sjuzet, the discursive organization through which themes are staged. We study this through the recurrence of literary motifs (blood in Macbeth, for instance), where symbolic significance emerges not from local cues, but through repetition-with-variation across the whole work. We operationalize this setting as Computational Motif Discovery, a transductive task that predicts missing line-motif links using the full narrative structure of a play. We propose Critic's Eye, which models the work as a heterogeneous Narrative Topology Graph and performs discriminative inference over global structural evidence. Critic's Eye achieves 84.8% Hit@5 on our benchmark, nearly quadrupling state-of-the-art proprietary foundation models (Gemini-3-pro-preview, ~22.6%). Our analysis reveals that despite massive scaling, sequence-based models hit a "semantic ceiling", struggling to resolve dispersed dependencies. These findings suggest that explicit structural priors are a far more effective inductive bias than parameter scaling for decoding the complex architecture of long-form literature.