A Long-Term Empirical Study of Aesthetic Differences in Female Characters of Chinese Animation: Quantitative Analysis Based on a Three-Tier Aesthetic Framework and Dual-Perspective Large Language Model Analysis

Abstract

Against the backdrop of digital media iteration and the expansion of China’s animation industry, aesthetic discourse surrounding female animated characters demonstrates differentiated structural features across production eras and evaluation contexts. Existing research is constrained by fragmented indicator systems, single-modality data, and a lack of standardized frameworks to support long-term cross-era quantitative comparison. To address these gaps, this study constructs a three-tier aesthetic evaluation framework consisting of Beauty of Image, Beauty of Impression, and Beauty of Idea, and proposes a dual-perspective quantitative workflow integrating large language model-based text mining, multimodal visual feature extraction, and the entropy-weighted TOPSIS method. Based on 72 representative Chinese animated works, 851,217 raw user comments, and 261 standardized character images, this study analyzes the distribution of aesthetic dimensions in both narrative-comment and pure-visual scenarios respectively, and validates the methodological reliability from three aspects: inter-annotator agreement, human–machine consistency, and test-retest stability. Quantitative results reveal that the two contexts present divergent aesthetic weight structures: identity-related dimensions exhibit the highest discriminative salience in narrative scenarios, while visual morphological attributes dominate in pure-visual settings. Across production eras, the discriminative salience of static morphological dimensions shows a steady upward trend, whereas the differentiating power of traditional cultural aesthetic dimensions declines gradually. High-frequency aesthetic vocabulary presents identity-oriented structural features in narrative contexts and form-and-style-oriented features in pure-visual scenarios respectively, with an overall evolutionary trend shifting toward visual detail orientation over time. This study empirically validates the effectiveness of the three-tier aesthetic framework and provides a standardized cross-era quantitative analytical paradigm for animation aesthetics research.

The authors' abstract, as published at the source. Applied Sciences, 2026 · DOI ↗