Semantic-Disentangled Transformer With Noun-Verb Embedding for Compositional Action Recognition | IEEE Journals & Magazine | IEEE Xplore