HAIR: Hierarchical Visual-Semantic Relational Reasoning for Video Question Answering | IEEE Conference Publication | IEEE Xplore