Word card generation for language education using latent diffusion model
J. -Y. Zhou, T. -H. Fu · IET conference proceedings. · 2024
This paper introduces an innovative system for word card generation in language education, utilizing the Latent Diffusion Model (LDM). The system addresses challenges of dealing with copyright compliance, semantic relevance, and inappropriate content in the generated word cards. By integrating CC0 datasets, GPT-4 prompt generation, and Safe Latent Diffusion (SLD) mechanisms for more secure image usage, the system generates contextually relevant word cards under copyright regulations. Experimental evaluations demonstrate the system's success in producing visually appealing word cards, while effectively mitigating the risk of inappropriate content at the same time.