Edit model card

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
ุฃู„ุฏูŠูƒ ู‚ู„ู… ุŸ ุฃู„ุฏูŠูƒ ู‚ู„ู…
ู„ูŠุณุช ู‡ู†ุงูƒ ู…ุณุงูุฉ ุนู„ู‰ ู‡ุฐู‡ ุงู„ุฃุฑุถ ุฃุจุนุฏ ู…ู† ูŠูˆู… ุฃู…ุณ. ู„ูŠุณุช ู†ุงู„ูƒ ู…ุณุงูุฉ ุนู„ู‰ ู‡ุฐู‡ ุงู„ุฃุฑุถ ุฃุจุนุฏ ู…ู† ูŠูˆู… ุงู„ุฃู…ุณ ู…
ุฅู†ูƒ ุชูƒุจุฑ ุงู„ู…ุดูƒู„ุฉ. ุฅู†ูƒ ุชูƒุจุฑ ุงู„ู…ุดูƒู„ุฉ
ูŠุฑุบุจ ุฃู† ูŠู„ุชู‚ูŠ ุจูƒ. ูŠุฑุบุจ ุฃู† ูŠู„ุชู‚ูŠ ุจูƒ
ุฅู†ู‡ู… ู„ุง ูŠุนุฑููˆู† ู„ู…ุงุฐุง ุญุชู‰. ุฅู†ู‡ู… ู„ุง ูŠุนุฑููˆู† ู„ู…ุงุฐุง ุญุชู‰
ุณูŠุณุนุฏู†ูŠ ู…ุณุงุนุฏุชูƒ ุฃูŠ ูˆู‚ุช ุชุญุจ. ุณูŠุณุฆุฏู†ูŠู…ุณุงุนุฏุชูƒ ุฃูŠ ูˆู‚ุฏ ุชุญุจ
ุฃูŽุญูŽุจู‘ู ู†ุธุฑูŠู‘ุฉ ุนู„ู…ูŠุฉ ุฅู„ูŠู‘ ู‡ูŠ ุฃู† ุญู„ู‚ุงุช ุฒุญู„ ู…ูƒูˆู†ุฉ ุจุงู„ูƒุงู…ู„ ู…ู† ุงู„ุฃู…ุชุนุฉ ุงู„ู…ูู‚ูˆุฏุฉ. ุฃุญุจ ู†ุธุฑูŠุฉ ุนู„ู…ูŠุฉ ุฅู„ูŠ ู‡ูŠ ุฃู† ุญู„ ู‚ุชุฒุญ ุงู„ู…ูƒูˆูŠู†ุง ุจุงู„ูƒุงู…ู„ ู…ู† ุงู„ุฃู…ุช ุนู† ุงู„ู…ูู‚ูˆุฏุฉ
ุณุฃุดุชุฑูŠ ู„ู‡ ู‚ู„ู…ุงู‹. ุณุฃุดุชุฑูŠ ู„ู‡ ู‚ู„ู…ุง
ุฃูŠู† ุงู„ู…ุดูƒู„ุฉ ุŸ ุฃูŠู† ุงู„ู…ุดูƒู„
ูˆูŽู„ูู„ู‘ูŽู‡ู ูŠูŽุณู’ุฌูุฏู ู…ูŽุง ูููŠ ุงู„ุณู‘ูŽู…ูŽุงูˆูŽุงุชู ูˆูŽู…ูŽุง ูููŠ ุงู„ู’ุฃูŽุฑู’ุถู ู…ูู†ู’ ุฏูŽุงุจู‘ูŽุฉู ูˆูŽุงู„ู’ู…ูŽู„ูŽุงุฆููƒูŽุฉู ูˆูŽู‡ูู…ู’ ู„ูŽุง ูŠูŽุณู’ุชูŽูƒู’ุจูุฑููˆู†ูŽ ูˆู„ู„ู‡ ูŠุณุฌุฏ ู…ุง ููŠ ุงู„ุณู…ุงูˆุงุช ูˆู…ุง ููŠ ุงู„ุฃุฑุถ ู…ู† ุฏุงุจุฉ ูˆุงู„ู…ู„ุงุฆูƒุฉ ูˆู‡ู… ู„ุง ูŠุณุชูƒุจุฑูˆู†

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", "แƒป", "~", "ีž",
                  "ุŸ", "ุŒ", "เฅค", "เฅฅ", "ยซ", "ยป", "โ€ž", "โ€œ", "โ€", "ใ€Œ", "ใ€", "โ€˜", "โ€™", "ใ€Š", "ใ€‹", "(", ")", "[", "]",
                  "{", "}", "=", "`", "_", "+", "<", ">", "โ€ฆ", "โ€“", "ยฐ", "ยด", "สพ", "โ€น", "โ€บ", "ยฉ", "ยฎ", "โ€”", "โ†’", "ใ€‚",
                  "ใ€", "๏น‚", "๏น", "โ€ง", "๏ฝž", "๏น", "๏ผŒ", "๏ฝ›", "๏ฝ", "๏ผˆ", "๏ผ‰", "๏ผป", "๏ผฝ", "ใ€", "ใ€‘", "โ€ฅ", "ใ€ฝ",
                  "ใ€Ž", "ใ€", "ใ€", "ใ€Ÿ", "โŸจ", "โŸฉ", "ใ€œ", "๏ผš", "๏ผ", "๏ผŸ", "โ™ช", "ุ›", "/", "\\", "ยบ", "โˆ’", "^", "'", "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
475,896

Datasets used to train jonatasgrosman/wav2vec2-large-xlsr-53-arabic

Spaces using jonatasgrosman/wav2vec2-large-xlsr-53-arabic 7

Evaluation results