Bp500 Base10k Voxpopuli

🚀 bp500-base10k_voxpopuli: Wav2vec 2.0 with Brazilian Portuguese (BP) Dataset

This project demonstrates a fine - tuned Wav2vec model for Brazilian Portuguese. It utilizes multiple datasets to enhance the model's performance in automatic speech recognition for Brazilian Portuguese.

🔍 Key Information

📊 Model Results

The model bp500 - base10k_voxpopuli was evaluated on the Common Voice dataset for the Automatic Speech Recognition task, achieving a Test WER of 24.9.

📦 Datasets Used

• CETUC: It contains about 145 hours of Brazilian Portuguese speech from 50 male and 50 female speakers. Each speaker pronounces around 1,000 phonetically balanced sentences selected from the CETEN - Folha corpus.
• Common Voice 7.0: A project by the Mozilla Foundation aiming to create an open - source dataset in multiple languages. Volunteers contribute and validate speech through the official site.
• [Lapsbm](https://github.com/falabrasil/gitlab - resources): Used by the Fala Brasil group to benchmark ASR systems in Brazilian Portuguese. It has 35 speakers (10 females), with each speaker pronouncing 20 unique sentences, totaling 700 utterances. The audios were recorded at 22.05 kHz without environment control.
• Multilingual Librispeech (MLS): A large - scale dataset available in many languages. The Portuguese subset (mostly Brazilian variant) used in this work has about 284 hours of speech from 55 audiobooks read by 62 speakers.
• Multilingual TEDx: A collection of audio recordings from TEDx talks in 8 source languages. The Portuguese set (mostly Brazilian Portuguese) contains 164 hours of transcribed speech.
• Sidney (SID): Contains 5,777 utterances from 72 speakers (20 women) aged 17 - 59, with additional information like place of birth, age, gender, education, and occupation.
• VoxForge: A project to build open datasets for acoustic models. The corpus has around 100 speakers and 4,130 utterances of Brazilian Portuguese, with sample rates ranging from 16kHz to 44.1kHz.

📊 Dataset Usage Statistics

📈 Model Summary

📝 Transcription Examples

💻 Usage Examples

Basic Usage

MODEL_NAME = "lgris/bp500-base10k_voxpopuli" 

Imports and Dependencies

%%capture
!pip install torch==1.8.2+cu111 torchvision==0.9.2+cu111 torchaudio===0.8.2 -f https://download.pytorch.org/whl/lts/1.8/torch_lts.html
!pip install datasets
!pip install jiwer
!pip install transformers
!pip install soundfile
!pip install pyctcdecode
!pip install https://github.com/kpu/kenlm/archive/master.zip

import jiwer
import torchaudio
from datasets import load_dataset, load_metric
from transformers import (
    Wav2Vec2ForCTC,
    Wav2Vec2Processor,
)
from pyctcdecode import build_ctcdecoder
import torch
import re
import sys

Helpers

chars_to_ignore_regex = '[\,\?\.\!\;\:\"]'  # noqa: W605

def map_to_array(batch):
    speech, _ = torchaudio.load(batch["path"])
    batch["speech"] = speech.squeeze(0).numpy() 
    batch["sampling_rate"] = 16_000 
    batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower().replace("’", "'")
    batch["target"] = batch["sentence"]
    return batch

def calc_metrics(truths, hypos):
    wers = []
    mers = []
    wils = []
    for t, h in zip(truths, hypos):
        try:
            wers.append(jiwer.wer(t, h))
            mers.append(jiwer.mer(t, h))
            wils.append(jiwer.wil(t, h))
        except: # Empty string?
            pass
    wer = sum(wers)/len(wers)
    mer = sum(mers)/len(mers)
    wil = sum(wils)/len(wils)
    return wer, mer, wil

def load_data(dataset):
    data_files = {'test': f'{dataset}/test.csv'}
    dataset = load_dataset('csv', data_files=data_files)["test"]
    return dataset.map(map_to_array)

Model

class STT:

    def __init__(self, 
                 model_name, 
                 device='cuda' if torch.cuda.is_available() else 'cpu', 
                 lm=None):
        self.model_name = model_name
        self.model = Wav2Vec2ForCTC.from_pretrained(model_name).to(device)
        self.processor = Wav2Vec2Processor.from_pretrained(model_name)
        self.vocab_dict = self.processor.tokenizer.get_vocab()
        self.sorted_dict = {
            k.lower(): v for k, v in sorted(self.vocab_dict.items(), 
                                            key=lambda item: item[1])
        }
        self.device = device
        self.lm = lm
        if self.lm:            
            self.lm_decoder = build_ctcdecoder(
                list(self.sorted_dict.keys()),
                self.lm
            )

    def batch_predict(self, batch):
        features = self.processor(batch["speech"], 
                                  sampling_rate=batch["sampling_rate"][0], 
                                  padding=True, 
                                  return_tensors="pt")
        input_values = features.input_values.to(self.device)
        with torch.no_grad():
            logits = self.model(input_values).logits
        if self.lm:
            logits = logits.cpu().numpy()
            batch["predicted"] = []
            for sample_logits in logits:
                batch["predicted"].append(self.lm_decoder.decode(sample_logits))
        else:
            pred_ids = torch.argmax(logits, dim=-1)
            batch["predicted"] = self.processor.batch_decode(pred_ids)
        return batch

Download Datasets

%%capture
!gdown --id 1HFECzIizf-bmkQRLiQD0QVqcGtOG5upI
!mkdir bp_dataset
!unzip bp_dataset -d bp_dataset/

%cd bp_dataset 

Tests

stt = STT(MODEL_NAME)

CETUC

ds = load_data('cetuc_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8) 
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("CETUC WER:", wer)

Common Voice

ds = load_data('commonvoice_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8) 
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("CV WER:", wer)

LaPS

ds = load_data('lapsbm_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8) 
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("Laps WER:", wer)

MLS

ds = load_data('mls_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8) 
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("MLS WER:", wer)

SID

ds = load_data('sid_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8) 
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("Sid WER:", wer)

TEDx

ds = load_data('tedx_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8) 
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("TEDx WER:", wer)

VoxForge

ds = load_data('voxforge_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8) 
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("VoxForge WER:", wer)

Tests with LM

!rm -rf ~/.cache
!gdown --id 1GJIKseP5ZkTbllQVgOL98R4yYAcIySFP  # trained with wikipedia
stt = STT(MODEL_NAME, lm='pt-BR-wiki.word.4-gram.arpa')
# !gdown --id 1dLFldy7eguPtyJj5OAlI4Emnx0BpFywg  # trained with bp
# stt = STT(MODEL_NAME, lm='pt-BR.word.4-gram.arpa')

Cetuc

ds = load_data('cetuc_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8) 
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("CETUC WER:", wer)

Common Voice

ds = load_data('commonvoice_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8) 
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("CV WER:", wer)

LaPS

ds = load_data('lapsbm_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8) 
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("Laps WER:", wer)

MLS

ds = load_data('mls_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8) 
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("MLS WER:", wer)

SID

ds = load_data('sid_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8) 
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("Sid WER:", wer)

TEDx

ds = load_data('tedx_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8) 
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("TEDx WER:", wer)

VoxForge

ds = load_data('voxforge_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8) 
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("VoxForge WER:", wer)

📄 License

This project is licensed under the [apache - 2.0](https://www.apache.org/licenses/LICENSE - 2.0) license.