Author(s)

Shuyang Yu

Corresponding Author

Shuyang Yu

Abstract

This paper proposes a data-driven multidimensional framework to evaluate cultural works across films, television series, and novels. The framework integrates character relationship networks, narrative structure and perspective, audience emotion and empathy, and fairness-aware representation indices into four comprehensive measures: Representation Equity, Narrative Richness, Character Attractiveness, and Audience Engagement. We construct a real multi-source dataset spanning 2018–2023 and benchmark three predictive models for a normalized success score. Experiments show that gradient boosting outperforms linear and bagging baselines on accuracy and fairness; audience engagement emerges as the strongest correlate of success, followed by character attractiveness and equitable representation, while narrative richness exhibits a moderate effect. Cross-media analysis demonstrates consistent performance across films, TV series, and novels, and ablation confirms the necessity of each feature family, especially audience engagement signals. The results validate that combining network science, NLP-based narrative analysis, and fairness metrics yields a practical, interpretable, and inclusive evaluation tool that links creative decisions to measurable outcomes and supports content optimization at scale.

Keywords

Cultural analytics, narrative modeling, fairness, success prediction