A Visually-grounded First-person Dialogue Dataset with Verbal and Non-verbal Responses

29688

Knowledge Triples