related documents CRAPES:Cross-modal Annotation Projection for Visual Semantic Role Labeling Conference Proceeding