README.md

metadata

language: es
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 Spanish by Jonatas Grosman
    results:
      - task:
          name: Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: Common Voice es
          type: common_voice
          args: es
        metrics:
          - name: Test WER
            type: wer
            value: 8.81
          - name: Test CER
            type: cer
            value: 2.7

Wav2Vec2-Large-XLSR-53-Spanish

Fine-tuned facebook/wav2vec2-large-xlsr-53 on Spanish 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("es", model_path="jonatasgrosman/wav2vec2-large-xlsr-53-spanish")

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 = "es"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-spanish"
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
HABITA EN AGUAS POCO PROFUNDAS Y ROCOSAS.	HABITAN AGUAS POCO PROFUNDAS Y ROCOSAS
OPERA PRINCIPALMENTE VUELOS DE CABOTAJE Y REGIONALES DE CARGA.	OPERA PRINCIPALMENTE VUELO DE CARBOTAJES Y REGIONALES DE CARGAN
PARA VISITAR CONTACTAR PRIMERO CON LA DIRECCIÓN.	PARA VISITAR CONTACTAR PRIMERO CON LA DIRECCIÓN
TRES	TRES
REALIZÓ LOS ESTUDIOS PRIMARIOS EN FRANCIA, PARA CONTINUAR LUEGO EN ESPAÑA.	REALIZÓ LOS ESTUDIOS PRIMARIOS EN FRANCIA PARA CONTINUAR LUEGO EN ESPAÑA
EN LOS AÑOS QUE SIGUIERON, ESTE TRABAJO ESPARTA PRODUJO DOCENAS DE BUENOS JUGADORES.	EN LOS AÑOS QUE SIGUIERON ESTE TRABAJO ESPARTA PRODUJO DOCENA DE BUENOS JUGADORES
SE ESTÁ TRATANDO DE RECUPERAR SU CULTIVO EN LAS ISLAS CANARIAS.	SE ESTÓ TRATANDO DE RECUPERAR SU CULTIVO EN LAS ISLAS CANARIAS
SÍ	SÍ
"FUE ""SACADA"" DE LA SERIE EN EL EPISODIO ""LEAD"", EN QUE ALEXANDRA CABOT REGRESÓ."	FUE SACADA DE LA SERIE EN EL EPISODIO LEED EN QUE ALEXANDRA KAOT REGRESÓ
SE UBICAN ESPECÍFICAMENTE EN EL VALLE DE MOKA, EN LA PROVINCIA DE BIOKO SUR.	SE UBICAN ESPECÍFICAMENTE EN EL VALLE DE MOCA EN LA PROVINCIA DE PÍOCOSUR

Evaluation

The model can be evaluated as follows on the Spanish 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 = "es"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-spanish"
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-spanish	8.81%	2.70%
pcuenq/wav2vec2-large-xlsr-53-es	10.55%	3.20%
facebook/wav2vec2-large-xlsr-53-spanish	16.99%	5.40%
mrm8488/wav2vec2-large-xlsr-53-spanish	19.20%	5.96%