File size: 4,971 Bytes
700c2fc 772a66e 700c2fc 772a66e dbaca8d 64ad055 ad1c5c0 772a66e ad1c5c0 772a66e 700c2fc 772a66e ad1c5c0 772a66e ad1c5c0 1f640d7 ad1c5c0 700c2fc 772a66e |
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 |
---
language: lg
datasets:
- mozilla-foundation/common_voice_7_0
metrics:
- wer
tags:
- audio
- automatic-speech-recognition
- common_voice
- hf-asr-leaderboard
- lg
- robust-speech-event
- speech
license: apache-2.0
model-index:
- name: Wav2Vec2 Luganda by Indonesian-NLP
results:
- task:
name: Speech Recognition
type: automatic-speech-recognition
dataset:
name: Common Voice lg
type: common_voice
args: lg
metrics:
- name: Test WER
type: wer
value: 9.332
- name: Test CER
type: cer
value: 1.987
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: Common Voice 7
type: mozilla-foundation/common_voice_7_0
args: lg
metrics:
- name: Test WER
type: wer
value: 13.844
- name: Test CER
type: cer
value: 2.68
---
# Automatic Speech Recognition for Luganda
This is the model built for the
[Mozilla Luganda Automatic Speech Recognition competition](https://zindi.africa/competitions/mozilla-luganda-automatic-speech-recognition).
It is a fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53)
model on the [Luganda Common Voice dataset](https://huggingface.co/datasets/common_voice) version 7.0.
We also provide a [live demo](https://huggingface.co/spaces/indonesian-nlp/luganda-asr) to test the model.
When using this model, make sure that your speech input is sampled at 16kHz.
## Usage
The model can be used directly (without a language model) as follows:
```python
import torch
import torchaudio
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
test_dataset = load_dataset("common_voice", "lg", split="test[:2%]")
processor = Wav2Vec2Processor.from_pretrained("indonesian-nlp/wav2vec2-luganda")
model = Wav2Vec2ForCTC.from_pretrained("indonesian-nlp/wav2vec2-luganda")
resampler = torchaudio.transforms.Resample(48_000, 16_000)
# Preprocessing the datasets.
# We need to read the aduio files as arrays
def speech_file_to_array_fn(batch):
if "audio" in batch:
speech_array = torch.tensor(batch["audio"]["array"])
else:
speech_array, sampling_rate = torchaudio.load(batch["path"])
batch["speech"] = resampler(speech_array).squeeze().numpy()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset[:2]["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
predicted_ids = torch.argmax(logits, dim=-1)
print("Prediction:", processor.batch_decode(predicted_ids))
print("Reference:", test_dataset[:2]["sentence"])
```
## Evaluation
The model can be evaluated as follows on the Indonesian test data of Common Voice.
```python
import torch
import torchaudio
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
import re
test_dataset = load_dataset("common_voice", "lg", split="test")
wer = load_metric("wer")
processor = Wav2Vec2Processor.from_pretrained("indonesian-nlp/wav2vec2-luganda")
model = Wav2Vec2ForCTC.from_pretrained("indonesian-nlp/wav2vec2-luganda")
model.to("cuda")
chars_to_ignore = [",", "?", ".", "!", "-", ";", ":", '""', "%", "'", '"', "�", "‘", "’", "’"]
chars_to_ignore_regex = f'[{"".join(chars_to_ignore)}]'
resampler = torchaudio.transforms.Resample(48_000, 16_000)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower()
if "audio" in batch:
speech_array = torch.tensor(batch["audio"]["array"])
else:
speech_array, sampling_rate = torchaudio.load(batch["path"])
batch["speech"] = resampler(speech_array).squeeze().numpy()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def evaluate(batch):
inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values.to("cuda"), attention_mask=inputs.attention_mask.to("cuda")).logits
pred_ids = torch.argmax(logits, dim=-1)
batch["pred_strings"] = processor.batch_decode(pred_ids)
return batch
result = test_dataset.map(evaluate, batched=True, batch_size=8)
print("WER: {:2f}".format(100 * wer.compute(predictions=result["pred_strings"], references=result["sentence"])))
```
WER without KenLM: 15.38 %
WER With KenLM:
**Test Result**: 7.53 %
## Training
The Common Voice `train`, `validation`, and ... datasets were used for training as well as ... and ... # TODO
The script used for training can be found [here](https://github.com/indonesian-nlp/luganda-asr)
|