metadata

language: pt
datasets:
  - common_voice
metrics:
  - wer
  - cer
tags:
  - audio
  - automatic-speech-recognition
  - speech
  - xlsr-fine-tuning-week
license: apache-2.0
model-index:
  - name: XLSR Wav2Vec2 Portuguese by Jonatas Grosman
    results:
      - task:
          name: Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: Common Voice pt
          type: common_voice
          args: pt
        metrics:
          - name: Test WER
            type: wer
            value: 11.62
          - name: Test CER
            type: cer
            value: 3.91

Wav2Vec2-Large-XLSR-53-Portuguese

Fine-tuned facebook/wav2vec2-large-xlsr-53 on Portuguese using the Common Voice. 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 ASRecognition library:

from asrecognition import ASREngine

asr = ASREngine("pt", model_path="jonatasgrosman/wav2vec2-large-xlsr-53-portuguese")

audio_paths = ["/path/to/file.mp3", "/path/to/another_file.wav"]
transcriptions = asr.transcribe(audio_paths)

Writing your own inference script:

import torch
import librosa
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor

LANG_ID = "pt"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-portuguese"
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
NEM O RADAR NEM OS OUTROS INSTRUMENTOS DETECTARAM O BOMBARDEIRO STEALTH.	NEMHUM VADAN OS OLTWES INSTRUMENTOS DE TTÉÃN UM BOMBERDEIRO OSTER
PEDIR DINHEIRO EMPRESTADO ÀS PESSOAS DA ALDEIA	E DIR ENGINHEIRO EMPRESTAR AS PESSOAS DA ALDEIA
OITO	OITO
TRANCÁ-LOS	TRANCAUVOS
REALIZAR UMA INVESTIGAÇÃO PARA RESOLVER O PROBLEMA	REALIZAR UMA INVESTIGAÇÃO PARA RESOLVER O PROBLEMA
O YOUTUBE AINDA É A MELHOR PLATAFORMA DE VÍDEOS.	YOUTUBE AINDA É A MELHOR PLATAFOMA DE VÍDEOS
MENINA E MENINO BEIJANDO NAS SOMBRAS	MENINA E MENINO BEIJANDO NAS SOMBRAS
EU SOU O SENHOR	EU SOU O SENHOR
DUAS MULHERES QUE SENTAM-SE PARA BAIXO LENDO JORNAIS.	DUAS MIERES QUE SENTAM-SE PARA BAICLANE JODNÓI
EU ORIGINALMENTE ESPERAVA	EU ORIGINALMENTE ESPERAVA

Evaluation

The model can be evaluated as follows on the Portuguese 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 = "pt"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-portuguese"
DEVICE = "cuda"

CHARS_TO_IGNORE = [",", "?", "¿", ".", "!", "¡", ";", "；", ":", '""', "%", '"', "�", "ʿ", "·", "჻", "~", "՞",
                   "؟", "،", "।", "॥", "«", "»", "„", "“", "”", "「", "」", "‘", "’", "《", "》", "(", ")", "[", "]",
                   "{", "}", "=", "`", "_", "+", "<", ">", "…", "–", "°", "´", "ʾ", "‹", "›", "©", "®", "—", "→", "。",
                   "、", "﹂", "﹁", "‧", "～", "﹏", "，", "｛", "｝", "（", "）", "［", "］", "【", "】", "‥", "〽",
                   "『", "』", "〝", "〟", "⟨", "⟩", "〜", "：", "！", "？", "♪", "؛", "/", "\\", "º", "−", "^", "ʻ", "ˆ"]

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 2021-04-22). 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-portuguese	11.62%	3.91%
joorock12/wav2vec2-large-xlsr-portuguese-a	15.52%	5.12%
joorock12/wav2vec2-large-xlsr-portuguese	15.95%	5.31%
gchhablani/wav2vec2-large-xlsr-pt	17.64%	6.04%
Rubens/Wav2Vec2-Large-XLSR-53-a-Portuguese	19.79%	6.57%
Rubens/Wav2Vec2-Large-XLSR-53-Portuguese	20.85%	7.08%
facebook/wav2vec2-large-xlsr-53-portuguese	26.73%	9.27%