A bandit approach to curriculum generation for automatic speech recognition
This work aims to improve ASR performance for low-resource languages by optimizing training data sequences, which is an incremental improvement for researchers in speech technology.
This paper addresses the challenge of low-data scenarios in Automated Speech Recognition (ASR), particularly for low-resource languages. The authors propose an Automated Curriculum Learning approach combined with an adversarial bandit framework to optimize the training sequence of mini-batches. Their method demonstrates an improvement over a baseline transfer-learning model when tested on a truly low-resource language.
The Automated Speech Recognition (ASR) task has been a challenging domain especially for low data scenarios with few audio examples. This is the main problem in training ASR systems on the data from low-resource or marginalized languages. In this paper we present an approach to mitigate the lack of training data by employing Automated Curriculum Learning in combination with an adversarial bandit approach inspired by Reinforcement learning. The goal of the approach is to optimize the training sequence of mini-batches ranked by the level of difficulty and compare the ASR performance metrics against the random training sequence and discrete curriculum. We test our approach on a truly low-resource language and show that the bandit framework has a good improvement over the baseline transfer-learning model.