Sensory Transfer Learning of Cultural Imagery by Integrating Text Sentiment Mining with Cross-regional Pattern Design

This paper proposes a sensory transfer learning method for cultural imagery that integrates text sentiment mining and cross-regional pattern design. Firstly, through text sentiment mining technology, emotional features are extracted from a large amount of text data to construct an emotional semantic space, so as to quantify the emotional expression of cultural images. Secondly, in combination with cross-regional pattern design, deep learning algorithms are utilized to extract and analyze the visual features of patterns from different regions, and a pattern feature library is established. On this basis, through the sensory transfer learning model, the emotional features of the text and the visual features of the patterns are fused to achieve cross-modal mapping of emotion and vision. The experimental results show that this new method can effectively capture the emotional connotation of cultural imagery and transfer it to cross-regional pattern design, providing a new idea and method for the design of cultural and creative products. This research not only enriches the theoretical system of perceptual design of cultural imagery, but also provides technical support for the development of cross-regional cultural and creative industries.