wav2vec2-large-xls-r-300m-basque Open-source Model - Free Implementation of Basque Automatic Speech Recognition

Wav2vec2 Large Xls R 300m Basque

Developed by deepdml

An automatic speech recognition model fine-tuned on the Basque Common Voice dataset based on facebook/wav2vec2-xls-r-300m

Speech Recognition

Transformers

OtherOpen Source License:Apache-2.0 #Basque speech recognition #Low character error rate (CER 10.01)#Multi-scenario speech transcription

Downloads 31

Release Time : 3/2/2022

Model Overview

This is an automatic speech recognition (ASR) model optimized for Basque, fine-tuned based on the wav2vec2-xls-r-300m architecture, suitable for Basque speech-to-text tasks.

Model Features

Optimized for Basque

Specifically fine-tuned for Basque, improving recognition accuracy for this language

Based on wav2vec2-xls-r architecture

Utilizes the powerful wav2vec2-xls-r-300m pre-trained model as its foundation, with excellent speech feature extraction capabilities

Trained on Common Voice dataset

Trained using the Mozilla Common Voice 7.0 Basque dataset, ensuring data diversity

Model Capabilities

Basque speech recognition

Speech-to-text

Automatic speech transcription

Use Cases

Speech transcription

Basque speech-to-text

Convert Basque speech content into text format

Word error rate 51.89%, character error rate 10.01%

Voice assistant

Basque voice interaction

Provide speech recognition capabilities for Basque voice assistants

🚀 wav2vec2-large-xls-r-300m-basque

This model is a fine - tuned version of facebook/wav2vec2-xls-r-300m on the common_voice dataset. It's designed for automatic speech recognition in Basque, offering valuable results on the evaluation set.

🚀 Quick Start

This model is a fine - tuned version of facebook/wav2vec2-xls-r-300m on the common_voice dataset. It achieves the following results on the evaluation set:

Loss: 0.4276
Wer: 0.5962

✨ Features

Metrics: Evaluated using WER (Word Error Rate) and CER (Character Error Rate).
Applicability: Suitable for automatic speech recognition tasks in Basque.
Data Source: Trained on the mozilla - foundation/common_voice_7_0 dataset.

📚 Documentation

Model Information

Property	Details
Model Type	wav2vec2-large-xls-r-300m-basque
Training Data	mozilla-foundation/common_voice_7_0
Metrics	WER, CER
Tags	automatic-speech-recognition, basque, generated_from_trainer, hf-asr-leaderboard, robust-speech-event
License	Apache-2.0

Model Index

Name: wav2vec2-large-xls-r-300m-basque
Results:
- Task:
  - Name: Automatic Speech Recognition
  - Type: automatic-speech-recognition
- Dataset:
  - Name: Common Voice 7
  - Type: mozilla-foundation/common_voice_7_0
  - Args: eu
- Metrics:
  - Name: Test WER
  - Type: wer
  - Value: 51.89
  - Name: Test CER
  - Type: cer
  - Value: 10.01

🔧 Technical Details

Training Procedure

Training Hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.0003
train_batch_size: 2
eval_batch_size: 2
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 4
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e - 08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 500
num_epochs: 5
mixed_precision_training: Native AMP

Training Results

Training Loss	Epoch	Step	Validation Loss	Wer
3.9902	1.29	400	2.1257	1.0
0.9625	2.59	800	0.5695	0.7452
0.4605	3.88	1200	0.4276	0.5962

Framework Versions

Transformers 4.16.2
Pytorch 1.10.0+cu111
Datasets 1.18.3
Tokenizers 0.11.0

📄 License

This project is licensed under the Apache - 2.0 license.

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご