metadata

language: nl
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 Dutch by Jonatas Grosman
    results:
      - task:
          name: Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: Common Voice nl
          type: common_voice
          args: nl
        metrics:
          - name: Test WER
            type: wer
            value: 15.76
          - name: Test CER
            type: cer
            value: 5.5

Wav2Vec2-Large-XLSR-53-Dutch

Fine-tuned facebook/wav2vec2-large-xlsr-53 on Dutch using the Common Voice and CSS10. When using this model, make sure that your speech input is sampled at 16kHz.

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:

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

LANG_ID = "nl"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-dutch"
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
DE ABORIGINALS ZIJN DE OORSPRONKELIJKE BEWONERS VAN AUSTRALIË.	DE ABBORIGENALS ZIJN DE OORSPRONKELIJKE BEWONERS VAN AUSTRALIË
MIJN TOETSENBORD ZIT VOL STOF.	MIJN TOETSENBORD ZIT VOL STOF
ZE HAD DE BANK BESCHADIGD MET HAAR SKATEBOARD.	ZE HAD DE BANK BESCHADIGD MET HAAR SCHEETBOORD
WAAR LAAT JIJ JE ONDERHOUD DOEN?	WAAR LAAT JIJ HET ONDERHOUD DOEN
NA HET LEZEN VAN VELE BEOORDELINGEN HAD ZE EINDELIJK HAAR OOG LATEN VALLEN OP EEN LAPTOP MET EEN QWERTY TOETSENBORD.	NA HET LEZEN VAN VELE BEOORDELINGEN HAD ZE EINDELIJK HAAR OOG LATEN VALLEN OP EEN LAPTOP MET EEN QUERTITOETSEMBORD
DE TAMPONS ZIJN OP.	DE TAPONT ZIJN OP
MARIJKE KENT OLIVIER NU AL MEER DAN TWEE JAAR.	MAARRIJKEN KENT OLIEVIER NU AL MEER DAN TWEE JAAR
HET VOEREN VAN BROOD AAN EENDEN IS EIGENLIJK ONGEZOND VOOR DE BEESTEN.	HET VOEREN VAN BEUROT AAN EINDEN IS EIGENLIJK ONGEZOND VOOR DE BEESTEN
PARKET MOET JE STOFZUIGEN, TEGELS MOET JE DWEILEN.	PARKET MOET JE STOF ZUIGEN MAAR TEGELS MOET JE DWEILEN
IN ONZE BUURT KENT IEDEREEN ELKAAR.	IN ONZE BUURT KENT IEDEREEN ELKAAR

Evaluation

The model can be evaluated as follows on the Dutch 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 = "nl"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-dutch"
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 2021-04-21). 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-dutch	15.76%	5.50%
wietsedv/wav2vec2-large-xlsr-53-dutch	16.78%	5.60%
facebook/wav2vec2-large-xlsr-53-dutch	20.97%	7.24%
nithinholla/wav2vec2-large-xlsr-53-dutch	21.39%	7.29%
MehdiHosseiniMoghadam/wav2vec2-large-xlsr-53-Dutch	25.89%	9.12%
simonsr/wav2vec2-large-xlsr-dutch	38.34%	13.29%