Are Language-and-Vision Transformers Sensitive to Discourse? A Case Study of ViLBERT

27289

Knowledge Triples

Source Entity	Relationship	Target Entity
CaptionGeneration Task	requires	Narrative Structure OtherEntity
Image Caption Matching Task	requires	Narrative Structure OtherEntity
ViLBERT Method	appliesTo	CaptionGeneration Task
ViLBERT Method	appliesTo	Image Caption Matching Task