A Quantitative Analysis of Image-Text Dynamics and Strategic Choices in Multimodal Advertising

Abstract

In the era of digital globalization, the paradigm of transnational advertising has shifted from simple linguistic conversion to sophisticated multimodal orchestration. This study explores the provocative question: Is traditional translation theory still relevant in an age of visual dominance? Utilizing DJI’s global social media advertisements as a case study, the research employs a mixed-methods approach grounded in Eco-translatology. A four-stage empirical design was implemented, involving a parallel corpus (N=202) and a technical pipeline that integrates GPT-4V and Sentence-BERT to quantify “modal contribution” through an ablation study paradigm. The findings reveal a pronounced “visual-dominance, textual-marginalization” logic: while image semantic contribution significantly and positively predicts communicative popularity, textual contribution exerts a significant negative effect. Crucially, hierarchical regression analysis demonstrates that after controlling for macro-ecological constraints—such as cultural distance and market GDP—translation strategy (domestication vs. foreignization) per se exerts no independent significant effect on communicative outcomes. These results challenge the traditional text-centric focus of translation studies, suggesting that the visual mode now bears the primary “semantic load.” Methodologically, this research offers a replicable computational paradigm for the digital humanities, providing data-driven insights for global marketing managers to shift strategic focus from textual micro-adjustments to upstream multimodal content design.