TutorialVQA

AbstractDespite the number of currently available datasets on video-question answering, there still remains a need for a dataset involving multi-step and non-factoid answers. Moreover, relying on video transcripts remains an …...