MooER: LLM-based Speech Recognition and Translation Models from Moore Threads
This addresses the problem of reducing data requirements for speech AI tasks, offering an incremental improvement in efficiency for researchers and developers.
The paper tackles speech recognition and translation by training LLM-based models on a 5000-hour pseudo-labeled dataset, achieving performance comparable to open-source models trained with much more data and outperforming others on a translation test with a BLEU score of 25.2.
In this paper, we present MooER, a LLM-based large-scale automatic speech recognition (ASR) / automatic speech translation (AST) model of Moore Threads. A 5000h pseudo labeled dataset containing open source and self collected speech data is used for training. We achieve performance comparable to other open source models trained with up to hundreds of thousands of hours of labeled speech data. Meanwhile, experiments conducted on Covost2 Zh2en testset suggest that our model outperforms other open source Speech LLMs. A BLEU score of 25.2 can be obtained. The main contributions of this paper are summarized as follows. First, this paper presents a training strategy for encoders and LLMs on speech related tasks (including ASR and AST) using a small size of pseudo labeled data without any extra manual annotation and selection. Second, we release our ASR and AST models and plan to open-source our training code and strategy in the near future. Moreover, a model trained on 8wh scale training data is planned to be released later on.