Fine-tuned XLSR-53 large model for speech recognition in Arabic
Fine-tuned facebook/wav2vec2-large-xlsr-53 on Arabic using the train and validation splits of Common Voice 6.1 and Arabic Speech Corpus. When using this model, make sure that your speech input is sampled at 16kHz.
This model has been fine-tuned thanks to the GPU credits generously given by the OVHcloud :)
The script used for training can be found here: https://github.com/jonatasgrosman/wav2vec2-sprint
Usage
The model can be used directly (without a language model) as follows...
Using the HuggingSound library:
from huggingsound import SpeechRecognitionModel
model = SpeechRecognitionModel("jonatasgrosman/wav2vec2-large-xlsr-53-arabic")
audio_paths = ["/path/to/file.mp3", "/path/to/another_file.wav"]
transcriptions = model.transcribe(audio_paths)
Writing your own inference script:
import torch
import librosa
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "ar"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-arabic"
SAMPLES = 10
test_dataset = load_dataset("common_voice", LANG_ID, split=f"test[:{SAMPLES}]")
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = batch["sentence"].upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["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)
predicted_sentences = processor.batch_decode(predicted_ids)
for i, predicted_sentence in enumerate(predicted_sentences):
print("-" * 100)
print("Reference:", test_dataset[i]["sentence"])
print("Prediction:", predicted_sentence)
Reference | Prediction |
---|---|
ุฃูุฏูู ููู ุ | ุฃูุฏูู ููู |
ููุณุช ููุงู ู ุณุงูุฉ ุนูู ูุฐู ุงูุฃุฑุถ ุฃุจุนุฏ ู ู ููู ุฃู ุณ. | ููุณุช ูุงูู ู ุณุงูุฉ ุนูู ูุฐู ุงูุฃุฑุถ ุฃุจุนุฏ ู ู ููู ุงูุฃู ุณ ู |
ุฅูู ุชูุจุฑ ุงูู ุดููุฉ. | ุฅูู ุชูุจุฑ ุงูู ุดููุฉ |
ูุฑุบุจ ุฃู ููุชูู ุจู. | ูุฑุบุจ ุฃู ููุชูู ุจู |
ุฅููู ูุง ูุนุฑููู ูู ุงุฐุง ุญุชู. | ุฅููู ูุง ูุนุฑููู ูู ุงุฐุง ุญุชู |
ุณูุณุนุฏูู ู ุณุงุนุฏุชู ุฃู ููุช ุชุญุจ. | ุณูุณุฆุฏููู ุณุงุนุฏุชู ุฃู ููุฏ ุชุญุจ |
ุฃูุญูุจูู ูุธุฑููุฉ ุนูู ูุฉ ุฅููู ูู ุฃู ุญููุงุช ุฒุญู ู ูููุฉ ุจุงููุงู ู ู ู ุงูุฃู ุชุนุฉ ุงูู ูููุฏุฉ. | ุฃุญุจ ูุธุฑูุฉ ุนูู ูุฉ ุฅูู ูู ุฃู ุญู ูุชุฒุญ ุงูู ููููุง ุจุงููุงู ู ู ู ุงูุฃู ุช ุนู ุงูู ูููุฏุฉ |
ุณุฃุดุชุฑู ูู ููู ุงู. | ุณุฃุดุชุฑู ูู ููู ุง |
ุฃูู ุงูู ุดููุฉ ุ | ุฃูู ุงูู ุดูู |
ููููููููู ููุณูุฌูุฏู ู ูุง ููู ุงูุณููู ูุงููุงุชู ููู ูุง ููู ุงููุฃูุฑูุถู ู ููู ุฏูุงุจููุฉู ููุงููู ูููุงุฆูููุฉู ููููู ู ููุง ููุณูุชูููุจูุฑูููู | ูููู ูุณุฌุฏ ู ุง ูู ุงูุณู ุงูุงุช ูู ุง ูู ุงูุฃุฑุถ ู ู ุฏุงุจุฉ ูุงูู ูุงุฆูุฉ ููู ูุง ูุณุชูุจุฑูู |
Evaluation
The model can be evaluated as follows on the Arabic test data of Common Voice.
import torch
import re
import librosa
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "ar"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-arabic"
DEVICE = "cuda"
CHARS_TO_IGNORE = [",", "?", "ยฟ", ".", "!", "ยก", ";", "๏ผ", ":", '""', "%", '"', "๏ฟฝ", "สฟ", "ยท", "แป", "~", "ี",
"ุ", "ุ", "เฅค", "เฅฅ", "ยซ", "ยป", "โ", "โ", "โ", "ใ", "ใ", "โ", "โ", "ใ", "ใ", "(", ")", "[", "]",
"{", "}", "=", "`", "_", "+", "<", ">", "โฆ", "โ", "ยฐ", "ยด", "สพ", "โน", "โบ", "ยฉ", "ยฎ", "โ", "โ", "ใ",
"ใ", "๏น", "๏น", "โง", "๏ฝ", "๏น", "๏ผ", "๏ฝ", "๏ฝ", "๏ผ", "๏ผ", "๏ผป", "๏ผฝ", "ใ", "ใ", "โฅ", "ใฝ",
"ใ", "ใ", "ใ", "ใ", "โจ", "โฉ", "ใ", "๏ผ", "๏ผ", "๏ผ", "โช", "ุ", "/", "\\", "ยบ", "โ", "^", "'", "สป", "ห"]
test_dataset = load_dataset("common_voice", LANG_ID, split="test")
wer = load_metric("wer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/wer.py
cer = load_metric("cer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/cer.py
chars_to_ignore_regex = f"[{re.escape(''.join(CHARS_TO_IGNORE))}]"
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
model.to(DEVICE)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
with warnings.catch_warnings():
warnings.simplefilter("ignore")
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = re.sub(chars_to_ignore_regex, "", batch["sentence"]).upper()
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(DEVICE), attention_mask=inputs.attention_mask.to(DEVICE)).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)
predictions = [x.upper() for x in result["pred_strings"]]
references = [x.upper() for x in result["sentence"]]
print(f"WER: {wer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
print(f"CER: {cer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
Test Result:
In the table below I report the Word Error Rate (WER) and the Character Error Rate (CER) of the model. I ran the evaluation script described above on other models as well (on 2021-05-14). Note that the table below may show different results from those already reported, this may have been caused due to some specificity of the other evaluation scripts used.
Model | WER | CER |
---|---|---|
jonatasgrosman/wav2vec2-large-xlsr-53-arabic | 39.59% | 18.18% |
bakrianoo/sinai-voice-ar-stt | 45.30% | 21.84% |
othrif/wav2vec2-large-xlsr-arabic | 45.93% | 20.51% |
kmfoda/wav2vec2-large-xlsr-arabic | 54.14% | 26.07% |
mohammed/wav2vec2-large-xlsr-arabic | 56.11% | 26.79% |
anas/wav2vec2-large-xlsr-arabic | 62.02% | 27.09% |
elgeish/wav2vec2-large-xlsr-53-arabic | 100.00% | 100.56% |
Citation
If you want to cite this model you can use this:
@misc{grosman2021xlsr53-large-arabic,
title={Fine-tuned {XLSR}-53 large model for speech recognition in {A}rabic},
author={Grosman, Jonatas},
howpublished={\url{https://huggingface.co/jonatasgrosman/wav2vec2-large-xlsr-53-arabic}},
year={2021}
}
- Downloads last month
- 1,268,516
Model tree for jonatasgrosman/wav2vec2-large-xlsr-53-arabic
Datasets used to train jonatasgrosman/wav2vec2-large-xlsr-53-arabic
Spaces using jonatasgrosman/wav2vec2-large-xlsr-53-arabic 19
Evaluation results
- Test WER on Common Voice arself-reported39.590
- Test CER on Common Voice arself-reported18.180