Open-source Speech Recognition Model english-filipino-wav2vec2-l-xls-r-test-05 - Supports English and Filipino Bilingual Speech-to-Text

English Filipino Wav2vec2 L Xls R Test 05

Developed by Khalsuu

This is a speech recognition model fine-tuned on Filipino speech datasets based on the wav2vec2-large-xlsr-53-english model, supporting English and Filipino speech-to-text tasks.

Speech Recognition

Transformers

Open Source License:Apache-2.0 #Filipino speech recognition #Multilingual speech transcription #Low word error rate

Downloads 67

Release Time : 5/6/2022

Model Overview

This model is specifically designed for automatic speech recognition (ASR) in English and Filipino, achieving good word error rate performance after fine-tuning on Filipino speech datasets.

Model Features

Cross-lingual speech recognition

Supports speech recognition in both English and Filipino

Based on XLSR pre-trained model

Built upon the powerful cross-lingual speech representation (XLSR) pre-trained model

Low word error rate

Achieved a word error rate of 26.84% on the evaluation set

Model Capabilities

English speech recognition

Filipino speech recognition

Speech-to-text

Use Cases

Speech transcription

Filipino speech transcription

Convert Filipino speech to text

Word error rate 26.84%

English speech transcription

Convert English speech to text

Good performance based on pre-trained model

🚀 english-filipino-wav2vec2-l-xls-r-test-05

This model is a fine - tuned version of jonatasgrosman/wav2vec2-large-xlsr-53-english on the filipino_voice dataset, which can be used for specific speech - related tasks.

🚀 Quick Start

This section provides a high - level overview of the model. For more detailed information, please refer to the following sections.

✨ Features

Fine - tuned on the filipino_voice dataset based on jonatasgrosman/wav2vec2-large-xlsr-53-english.
Achieved specific loss and WER results on the evaluation set.

📚 Documentation

📋 Model description

This model is a fine - tuned version of jonatasgrosman/wav2vec2-large-xlsr-53-english on the filipino_voice dataset. It achieves the following results on the evaluation set:

Loss: 0.4738
Wer: 0.2684

⚙️ Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.0003
train_batch_size: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 16
optimizer: Adam with betas=(0.9,0.999) and epsilon = 1e - 08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 500
num_epochs: 30
mixed_precision_training: Native AMP

📈 Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.3328	2.09	400	2.2174	0.9733
0.6432	4.19	800	0.3735	0.3896
0.2741	6.28	1200	0.3639	0.3425
0.1877	8.38	1600	0.3506	0.3425
0.1408	10.47	2000	0.3644	0.3181
0.1133	12.57	2400	0.3837	0.3047
0.0953	14.66	2800	0.4415	0.3103
0.0814	16.75	3200	0.3940	0.3092
0.0707	18.85	3600	0.4164	0.3013
0.059	20.94	4000	0.4488	0.2983
0.0545	23.04	4400	0.4803	0.3028
0.0482	25.13	4800	0.4731	0.2811
0.0426	27.23	5200	0.4606	0.2757
0.0395	29.32	5600	0.4738	0.2684

📊 Intended uses & limitations

More information needed

📊 Training and evaluation data

More information needed

🛠️ Training procedure

Framework versions

Transformers 4.11.3
Pytorch 1.10.0+cu113
Datasets 1.18.3
Tokenizers 0.10.3

📄 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