A Comparative Study of Visual and Multimodal Features in Critically Acclaimed vs. Popular Chinese Films Using the CMM Dataset

Abstract

The study revealed the multimodal differences between critically acclaimed and commercially successful Chinese films using the Chinese Multimodal Movie (CMM) dataset, comprising 781 films released between 2017 and 2024. Through computational analysis of visual, textual, temporal, and audience engagement features, the research identified distinct aesthetic and narrative strategies underpinning the two categories. Critically acclaimed films exhibited higher textual richness, greater lexical diversity in reviews, and longer narrative development windows, as evidenced by delayed peak box office days and sentimentally complex user engagement. Commercially successful films, in contrast, prioritized high brightness and visual vibrancy, concentrated early revenue, and aligned their release schedules with holidays and weekends to maximize audience turnout. Visual metrics, though individually non-discriminative, contributed meaningfully when fused with other modalities. Multimodal classification models trained on standardized features achieved over 93% accuracy, with plot embeddings and team influence scores emerging as top predictors for critical acclaim, while visual and temporal cues better forecasted commercial performance. The results revealed that aesthetic coherence and symbolic density typified critical recognition, whereas immediacy and sensory appeal underpinned mass-market viability. The findings bridge empirical media analysis with cultural discourse on cinematic value and audience behavior, offering a scalable framework for future multimodal film studies.