Affective-Computing-Driven Personalized Display of Cultural Information for Commercial Heritage Architecture

The display methods for traditional cultural heritage lack personalization and emotional interaction, making it difficult to stimulate the public’s deep cultural awareness. This is especially true in commercialized historical districts, where cultural value is easily overlooked. Balancing cultural value and commercial value in information display has become one of the challenges that needs to be addressed. To solve the above problems, this article focuses on the identification of deep cultural values and the optimization of the information display in Beijing’s Qianmen Street, proposing a framework for cultural information mining and display based on affective computing and large language models. The pre-trained models QwenLM and RoBERTa were employed to analyze text and image data from user-generated content on social media, identifying users’ emotional tendencies toward various cultural value dimensions and quantifying their multilayered understanding of architectural heritage. This study further constructed a multimodal information presentation model driven by emotional feedback, mapping it into virtual reality environments to enable personalized, multilayered cultural information visualization. The framework’s effectiveness was validated through an eye-tracking experiment that assessed how different presentation styles impacted users’ emotional engagement and cognitive outcomes. The results show that the affective computing and multimodal data fusion approach to cultural heritage presentation accurately captures users’ emotions, enhancing their interest and emotional involvement. Personalized presentations of information significantly improve users’ engagement, historical understanding, and cultural experience, thereby fostering a deeper comprehension of historical contexts and architectural details.