Driven digital preservation and rural revitalization: a multimodal framework integrating GANs and NLP for intangible cultural heritage

This paper studies the application possibilities of AI technology in promoting the integrated development of intangible cultural heritage and rural creativity, and focuses on the practicality of convolutional neural networks (CNN), generative adversarial networks (GAN), and Transformer in image recognition, generation, and text processing. By constructing a multimodal framework that integrates images, texts and sounds, an extensible solution suitable for multiple scenarios has been formed. In the Suzhou embroidery image recognition experiment, the accuracy rate of CNN is 91.4%. In the diffusion simulation process for Miao embroidery patterns, more than 2,000 samples were generated, among which 67% of the generated samples were actually applied. The click-through rate of the experimental platform using the recommendation system increased by 70.5%, and the conversion rate rose to 18.9%. It can be seen from this that AI can not only significantly facilitate the digital presentation of intangible cultural heritage information and effectively enhance the effect of dissemination and conversion, but also play a significant role in the construction of cultural and rural creative ecosystems, providing theoretical support and operational guidance for the establishment of an intelligent mechanism for the intangible cultural heritage industry.