Language Models as Few-Shot Learner for Task-Oriented Dialogue Systems

Task-oriented dialogue systems use four connected modules, namely, Natural Language Understanding (NLU), a Dialogue State Tracking (DST), Dialogue Policy (DP) and Natural Language Generation (NLG). A research challenge is to learn each module with the least amount of samples (i.e., few-shots) given the high cost related to the data collection... (read more)

Results in Papers With Code
(↓ scroll down to see all results)