Cang Jie System is an experimental project that aims to discuss the interrelations between the Chinese language, characters, and the algorithmic understanding of the meanings embedded in the characters. The project takes a technological approach to answer the question: how do Chinese character stores information in its form and shapes? In the light of a comprehensive study of Chinese characters, I proposed to create novel Chinese characters with generative deep neural networks and offered fresh perceptions of the character system by visually exploring the learnt latent space. In addition, the multi-media presentations of the experiment outcomes, including a live-generation web application and a data-driven virtual reality (VR) installation, were created to better communicate the experiment results to the mass audience.
In this project, I built a Variational Autoencoder – Generative Adversarial Networks (VAE-GAN) that could produce new characters based on existing characters and obtained thorough interpretations of the entire character system by exploring the latent space. My practice in constructing the VR installation also allowed me to investigate the potential of three-dimensional data visualization in VR environments. Cang Jie System inspires the audience to reflect upon the accepted written language system and ponder over the process of people documenting and communicating information with the assorted compositions of characters. It also contributes to the artificial intelligence (AI) art community and brings more insights to the integrated practice where AI meets art.