Online Continual Learning of End-to-End Speech Recognition Models
Online Continual Learning of End-to-End Speech Recognition Models
复制标题
端到端语音识别模型的在线持续学习
DOI：
--
复制
发表时间：
2022
期刊：
影响因子：
--
通讯作者：
Shinji Watanabe
中科院分区：
文献类型：
--
作者：
Muqiao Yang;Ian Lane;Shinji Watanabe
持续学习，也称为终身学习，旨在不断地从新数据中学习。虽然先前的研究持续学习的自动语音识别的重点是适应模型在多个不同的语音识别任务，在本文中，我们提出了一个实验设置， extit{在线连续学习}用于单个任务的自动语音识别。特别关注的情况下，额外的训练数据为同一个任务随着时间的推移逐渐变得可用，我们证明了执行增量模型更新的有效性，端到端的语音识别模型与在线梯度情景记忆（GEM）方法。此外，我们表明，通过在线持续学习和选择性采样策略，我们可以保持类似于从头开始重新训练模型的准确性，同时需要显着降低计算成本。我们还验证了我们的方法与自我监督学习（SSL）功能。
Continual Learning, also known as Lifelong Learning, aims to continually learn from new data as it becomes available. While prior research on continual learning in automatic speech recognition has focused on the adaptation of models across multiple different speech recognition tasks, in this paper we propose an experimental setting for extit{online continual learning} for automatic speech recognition of a single task. Specifically focusing on the case where additional training data for the same task becomes available incrementally over time, we demonstrate the effectiveness of performing incremental model updates to end-to-end speech recognition models with an online Gradient Episodic Memory (GEM) method. Moreover, we show that with online continual learning and a selective sampling strategy, we can maintain an accuracy that is similar to retraining a model from scratch while requiring significantly lower computation costs. We have also verified our method with self-supervised learning (SSL) features.