Personalized Generation and Effect Evaluation of Creative Writing of Marketing Communication Driven by Affective Computing

Abstract

In the digital age, traditional marketing communication increasingly struggles to address heterogeneous and rapidly changing individual needs. Affective computing provides new technical means for understanding and responding to user emotions, thereby enabling more precise and engaging creative writing in marketing contexts. This paper investigates the personalized generation and systematic effect evaluation of marketing communication texts driven by affective computing. First, a closed-loop framework is constructed that integrates multi-source emotional data collection, user emotional demand modeling, hierarchical content generation based on natural language generation techniques, and multi-channel emotional adaptation. Within this framework, emotional signals from diverse data sources are fused to build user profiles that guide the tailoring of tone, style, and narrative strategies in creative writing. Second, an evaluation model is proposed along the dimensions of emotion, behavior, and cognition to assess the effectiveness of generated content, including emotional resonance, engagement, and persuasive impact. Feedback from these indicators is used to iteratively optimize both the affective models and the generation strategies, forming an intelligent optimization loop. Furthermore, the study discusses key challenges such as data privacy, algorithmic bias, and content homogenization, and emphasizes the need for transparency and user consent in emotional data processing. Finally, it outlines future directions in multimodal fusion, human–machine collaboration, and intelligent evaluation mechanisms, aiming to provide theoretical and methodological support for enhancing the emotional impact, personalization, and accuracy of marketing communication content.