A Multimodal Feature Fusion Strategy for Cultural Authenticity Computing and Communication Optimization in Short Videos

Short videos, as a product of the deep integration of algorithmic recommendation and computer vision technologies, have become an important carrier for the dissemination of traditional culture. However, their fragmented communication mechanism often leads to the simplification and distortion of cultural content. This study proposes a Cultural Authenticity computing framework based on multimodal feature fusion and algorithmic optimization techniques. Through a controlled experiment comparing “authenticity-oriented” and “symbolization-oriented” short videos featuring the Baoxiang flower pattern, audience feedback was measured across cognitive, emotional, and behavioral dimensions. The experiment utilized structured cultural semantic embedding technology and timestamp annotation algorithms to construct authenticity-enhanced videos, using symbolization-oriented videos optimized based on an attention mechanism as the control. Results demonstrate that authenticity-enhanced videos significantly improve cultural understanding, strengthen emotional resonance, increase sharing willingness, and effectively reduce cultural misinterpretation. This study provides a technical pathway and empirical support for introducing cultural authenticity features into content-aware recommendation algorithms on short video platforms, achieving synergistic optimization between cultural communication effectiveness and technological implementation pathways.