Adaptive Endpointing with Deep Contextual Multi-armed Bandits
This work addresses adaptive endpointing for speech processing systems, offering an incremental improvement by replacing supervised learning with online bandit-based optimization.
The paper tackles the problem of adaptive endpointing by proposing a deep contextual multi-armed bandit method that selects optimal configurations online without ground truth labels or hyperparameter grid-search, resulting in reduced early cutoff errors while maintaining low latency.
Current endpointing (EP) solutions learn in a supervised framework, which does not allow the model to incorporate feedback and improve in an online setting. Also, it is a common practice to utilize costly grid-search to find the best configuration for an endpointing model. In this paper, we aim to provide a solution for adaptive endpointing by proposing an efficient method for choosing an optimal endpointing configuration given utterance-level audio features in an online setting, while avoiding hyperparameter grid-search. Our method does not require ground truth labels, and only uses online learning from reward signals without requiring annotated labels. Specifically, we propose a deep contextual multi-armed bandit-based approach, which combines the representational power of neural networks with the action exploration behavior of Thompson modeling algorithms. We compare our approach to several baselines, and show that our deep bandit models also succeed in reducing early cutoff errors while maintaining low latency.