Massively Multilingual ASR on 70 Languages: Tokenization, Architecture, and Generalization Capabilities
This work addresses the challenge of building efficient multilingual ASR systems for diverse languages, offering incremental improvements in performance and generalization.
The paper tackled scaling multilingual automatic speech recognition to 70 languages by exploring tokenization strategies and model architectures, achieving a 13.9%-15.6% average word error rate improvement over monolingual models and demonstrating generalization with 9.5% and 7.5% WER on an unseen dataset.
End-to-end multilingual ASR has become more appealing because of several reasons such as simplifying the training and deployment process and positive performance transfer from high-resource to low-resource languages. However, scaling up the number of languages, total hours, and number of unique tokens is not a trivial task. This paper explores large-scale multilingual ASR models on 70 languages. We inspect two architectures: (1) Shared embedding and output and (2) Multiple embedding and output model. In the shared model experiments, we show the importance of tokenization strategy across different languages. Later, we use our optimal tokenization strategy to train multiple embedding and output model to further improve our result. Our multilingual ASR achieves 13.9%-15.6% average WER relative improvement compared to monolingual models. We show that our multilingual ASR generalizes well on an unseen dataset and domain, achieving 9.5% and 7.5% WER on Multilingual Librispeech (MLS) with zero-shot and finetuning, respectively.