Summary of Mooer: Llm-based Speech Recognition and Translation Models From Moore Threads, by Junhao Xu et al.
MooER: LLM-based Speech Recognition and Translation Models from Moore Threads
by Junhao Xu, Zhenlin Liang, Yi Liu, Yichao Hu, Jian Li, Yajun Zheng, Meng Cai, Hua Wang
First submitted to arxiv on: 9 Aug 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper presents MooER, a large-scale automatic speech recognition (ASR) / automatic speech translation (AST) model based on LLMs of Moore Threads. The authors train their model using a pseudo-labeled dataset containing 5000 hours of open-source and self-collected speech data, achieving performance comparable to other open-source models trained with hundreds of thousands of hours of labeled data. Experiments on the Covost2 Zh2en testset show that MooER outperforms other open-source Speech LLMs, with a BLEU score of 25.2. The paper’s main contributions include a training strategy for encoders and LLMs on speech-related tasks using small pseudo-labeled datasets without manual annotation or selection. The authors also plan to release their ASR and AST models, as well as their training code and strategy. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary MooER is a new automatic speech recognition (ASR) / automatic speech translation (AST) model that uses large language models (LLMs). It’s like a superpower for computers that can understand and translate what people are saying. The researchers used a special kind of dataset called pseudo-labeled data to train their model, which allowed them to teach it without needing lots of human-annotated speech examples. They tested MooER against other open-source models trained with much more data, and it performed just as well! In fact, it even outperformed some of the other models on a specific test set. The researchers think this is because their training strategy allows for faster learning and better adaptation to new sounds. |
Keywords
» Artificial intelligence » Bleu » Translation
