A Generative AI Application for Qualitative Automatic Population of Multimedia Knowledge Graphs

Summary

The authors propose an approach for automatically populating a multimedia knowledge graph (MKG) using generative AI image synthesis. Their method computes a quality index (QI) from established evaluation metrics to replace much of the manual image quality checking, while keeping expert review for unclear or safety-critical cases. They built an MKG from its textual counterpart and report that QI-based rankings track expert judgments, evaluate faster, and match original training performance when neural networks are trained on the synthetic MKG.