| --- |
| language: yue |
| license: mit |
| library_name: transformers |
| pipeline_tag: automatic-speech-recognition |
| base_model: openai/whisper-large-v3 |
| tags: [automatic-speech-recognition, whisper, cantonese, buzzasr] |
| datasets: [google/fleurs] |
| metrics: [cer, wer] |
| --- |
| |
| # BuzzASR — Cantonese |
|
|
| A monolingual automatic speech recognition model for **Cantonese**, fine-tuned from |
| [openai/whisper-large-v3](https://huggingface.co/openai/whisper-large-v3). Part of **BuzzASR**, |
| a suite of 102 language-specialized ASR models (Findings of EMNLP 2026). |
|
|
| This model uses **simple fine-tuning (Whisper's tokenizer, ASR fine-tuning only)**. |
|
|
| > 🏆 **State-of-the-art (open-source).** On the combined FLEURS + Common Voice test set, this model |
| > achieves the lowest CER of every open system we compare against: Whisper-large-v3, Omnilingual 1B/7B, MMS, Qwen3-ASR, and Cohere Transcribe. |
|
|
| ## Results (normalized CER / WER, %) |
|
|
| | Test set | CER | WER | Whisper-large-v3 (zero-shot) CER | |
| |---|---|---|---| |
| | FLEURS | 21.29 | 39.22 | 58.02 | |
| | Common Voice 25 | 9.62 | 50.94 | 44.11 | |
| | Combined | 13.01 | 47.53 | 36.44 | |
|
|
| ~2.8x CER reduction over Whisper zero-shot on the combined test set. |
|
|
| ## Usage |
|
|
| ```python |
| import torch, torchaudio |
| from transformers import WhisperForConditionalGeneration, WhisperProcessor |
| |
| model = WhisperForConditionalGeneration.from_pretrained("BuzzASR/cantonese", torch_dtype=torch.float16).to("cuda").eval() |
| proc = WhisperProcessor.from_pretrained("BuzzASR/cantonese") |
| |
| wav, sr = torchaudio.load("audio.wav") # 16 kHz mono |
| feats = proc(wav[0], sampling_rate=16000, return_tensors="pt").input_features.to("cuda").half() |
| ids = model.generate(feats, num_beams=1, no_repeat_ngram_size=3, repetition_penalty=1.2) |
| print(proc.batch_decode(ids, skip_special_tokens=True)[0]) |
| ``` |
| The language/task prompt is baked into the generation config, so no `language=` argument is needed. |
|
|
| ## Training data |
| [FLEURS](https://huggingface.co/datasets/google/fleurs) + **Common Voice Corpus 25.0** (Mozilla, March 2025; https://commonvoice.mozilla.org/en/datasets), capped per the paper. Text-only data from the **Goldfish** corpus (Chang et al., 2026). |
|
|
| ## Limitations |
| Monolingual (Cantonese only). Evaluated on FLEURS / Common Voice test splits; other domains or dialects may differ. |
|
|
| ## Citation |
| Project page: https://lemn-lab.github.io/buzzasr-docs/ |
|
|
