File size: 7,365 Bytes
8aaca4c e8a247f 8aaca4c e8a247f ed50050 e8a247f 8aaca4c e8a247f e12deb1 ed50050 e12deb1 e8a247f 8aaca4c e8a247f 8aaca4c 11da310 8aaca4c 11da310 8aaca4c 49c3fc3 11da310 8aaca4c f3b683f 8aaca4c ce95bf7 8aaca4c e8a247f |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 |
---
language: vie
datasets:
- legacy-datasets/common_voice
- vlsp2020_vinai_100h
- AILAB-VNUHCM/vivos
- doof-ferb/vlsp2020_vinai_100h
- doof-ferb/fpt_fosd
- doof-ferb/infore1_25hours
- linhtran92/viet_bud500
- doof-ferb/LSVSC
- doof-ferb/vais1000
- doof-ferb/VietMed_labeled
- NhutP/VSV-1100
- doof-ferb/Speech-MASSIVE_vie
- doof-ferb/BibleMMS_vie
- capleaf/viVoice
metrics:
- wer
pipeline_tag: automatic-speech-recognition
tags:
- transcription
- audio
- speech
- chunkformer
- asr
- automatic-speech-recognition
license: cc-by-nc-4.0
model-index:
- name: ChunkFormer Large Vietnamese
results:
- task:
name: Speech Recognition
type: automatic-speech-recognition
dataset:
name: common-voice-vietnamese
type: common_voice
args: vi
metrics:
- name: Test WER
type: wer
value: 6.66
source:
name: Common Voice Vi Leaderboard
url: https://paperswithcode.com/sota/speech-recognition-on-common-voice-vi
- task:
name: Speech Recognition
type: automatic-speech-recognition
dataset:
name: VIVOS
type: vivos
args: vi
metrics:
- name: Test WER
type: wer
value: 4.18
source:
name: Vivos Leaderboard
url: https://paperswithcode.com/sota/speech-recognition-on-vivos
- task:
name: Speech Recognition
type: automatic-speech-recognition
dataset:
name: VLSP - Task 1
type: vlsp
args: vi
metrics:
- name: Test WER
type: wer
value: 14.09
---
# **ChunkFormer-Large-Vie: Large-Scale Pretrained ChunkFormer for Vietnamese Automatic Speech Recognition**
<style>
img {
display: inline;
}
</style>
[](https://paperswithcode.com/sota/speech-recognition-on-common-voice-vi?p=chunkformer-masked-chunking-conformer-for)
[](https://paperswithcode.com/sota/speech-recognition-on-vivos?p=chunkformer-masked-chunking-conformer-for)
[](https://creativecommons.org/licenses/by-nc/4.0/)
[](https://github.com/khanld/chunkformer)
[](https://arxiv.org/abs/2502.14673)
[](#description)
**!!!ATTENTION: Input audio must be MONO (1 channel) at 16,000 sample rate**
---
## Table of contents
1. [Model Description](#description)
2. [Documentation and Implementation](#implementation)
3. [Benchmark Results](#benchmark)
4. [Usage](#usage)
6. [Citation](#citation)
7. [Contact](#contact)
---
<a name = "description" ></a>
## Model Description
**ChunkFormer-Large-Vie** is a large-scale Vietnamese Automatic Speech Recognition (ASR) model based on the **ChunkFormer** architecture, introduced at **ICASSP 2025**. The model has been fine-tuned on approximately **3000 hours** of public Vietnamese speech data sourced from diverse datasets. A list of datasets can be found [**HERE**](dataset.tsv).
**!!! Please note that only the \[train-subset\] was used for tuning the model.**
---
<a name = "implementation" ></a>
## Documentation and Implementation
The [Documentation]() and [Implementation](https://github.com/khanld/chunkformer) of ChunkFormer are publicly available.
---
<a name = "benchmark" ></a>
## Benchmark Results
We evaluate the models using **Word Error Rate (WER)**. To ensure consistency and fairness in comparison, we manually apply **Text Normalization**, including the handling of numbers, uppercase letters, and punctuation.
1. **Public Models**:
| STT | Model | #Params | Vivos | Common Voice | VLSP - Task 1 | Avg. |
|-----|------------------------------------------------------------------------|---------|-------|--------------|---------------|------|
| 1 | **ChunkFormer** | 110M | 4.18 | 6.66 | 14.09 | **8.31** |
| 2 | [vinai/PhoWhisper-large](https://huggingface.co/vinai/PhoWhisper-large) | 1.55B | 4.67 | 8.14 | 13.75 | 8.85 |
| 3 | [nguyenvulebinh/wav2vec2-base-vietnamese-250h](https://huggingface.co/nguyenvulebinh/wav2vec2-base-vietnamese-250h) | 95M | 10.77 | 18.34 | 13.33 | 14.15 |
| 4 | [openai/whisper-large-v3](https://huggingface.co/openai/whisper-large-v3) | 1.55B | 8.81 | 15.45 | 20.41 | 14.89 |
| 5 | [khanhld/wav2vec2-base-vietnamese-160h](https://huggingface.co/khanhld/wav2vec2-base-vietnamese-160h) | 95M | 15.05 | 10.78 | 31.62 | 19.16 |
| 6 | [homebrewltd/Ichigo-whisper-v0.1](https://huggingface.co/homebrewltd/Ichigo-whisper-v0.1) | 22M | 13.46 | 23.52 | 21.64 | 19.54 |
2. **Private Models (API)**:
| STT | Model | VLSP - Task 1 |
|-----|--------|---------------|
| 1 | **ChunkFormer** | **14.1** |
| 2 | Viettel | 14.5 |
| 3 | Google | 19.5 |
| 4 | FPT | 28.8 |
---
<a name = "usage" ></a>
## Quick Usage
To use the ChunkFormer model for Vietnamese Automatic Speech Recognition, follow these steps:
1. **Download the ChunkFormer Repository**
```bash
git clone https://github.com/khanld/chunkformer.git
cd chunkformer
pip install -r requirements.txt
```
2. **Download the Model Checkpoint from Hugging Face**
```bash
pip install huggingface_hub
huggingface-cli download khanhld/chunkformer-large-vie --local-dir "./chunkformer-large-vie"
```
or
```bash
git lfs install
git clone https://huggingface.co/khanhld/chunkformer-large-vie
```
This will download the model checkpoint to the checkpoints folder inside your chunkformer directory.
3. **Run the model**
```bash
python decode.py \
--model_checkpoint path/to/local/chunkformer-large-vie \
--long_form_audio path/to/audio.wav \
--total_batch_duration 14400 \ #in second, default is 1800
--chunk_size 64 \
--left_context_size 128 \
--right_context_size 128
```
Example Output:
```
[00:00:01.200] - [00:00:02.400]: this is a transcription example
[00:00:02.500] - [00:00:03.700]: testing the long-form audio
```
**Advanced Usage** can be found [HERE](https://github.com/khanld/chunkformer/tree/main?tab=readme-ov-file#usage)
---
<a name = "citation" ></a>
## Citation
If you use this work in your research, please cite:
```bibtex
@inproceedings{chunkformer,
title={ChunkFormer: Masked Chunking Conformer For Long-Form Speech Transcription},
author={Khanh Le, Tuan Vu Ho, Dung Tran and Duc Thanh Chau},
booktitle={ICASSP},
year={2025}
}
```
---
<a name = "contact"></a>
## Contact
- [email protected]
- [](https://github.com/khanld)
- [](https://www.linkedin.com/in/khanhld257/) |