ICC: Quantifying Image Caption Concreteness for Multimodal Dataset
Curation
ICC: Quantifying Image Caption Concreteness for Multimodal Dataset
Curation
Web-scale training on paired text-image data is becoming increasingly central to multimodal learning, but is challenged by the highly noisy nature of datasets in the wild. Standard data filtering approaches succeed in removing mismatched text-image pairs, but permit semantically related but highly abstract or subjective text. These approaches lack the …