Zero-Shot Learning of Language Models for Describing Human Actions Based on Semantic Compositionality of Actions

Hideki Asoh, Ichiro Kobayashi