BootTOD proposes a self-bootstrapping framework to align context and response representations, improving task-oriented dialogue understanding.
BootTOD proposes a self-bootstrapping framework for task-oriented dialogue representations, outperforming contrastive methods.