wav2vec2-large-xls-r-300m-kurdish Open-source Model - Free Speech Recognition for Kurmanji Kurdish

Wav2vec2 Large Xls R 300m Kurdish

Developed by infinitejoy

This is an automatic speech recognition (ASR) model fine-tuned on Kurmanji Kurdish datasets based on facebook/wav2vec2-xls-r-300m.

Speech Recognition

Transformers

OtherOpen Source License:Apache-2.0 #Kurmanji speech recognition #Low-resource language ASR #XLS-R fine-tuning

Downloads 81

Release Time : 3/2/2022

Model Overview

This model is specifically designed for speech recognition tasks in the Kurmanji Kurdish language, fine-tuned on the Common Voice 7 dataset.

Model Features

Kurmanji Kurdish support

Speech recognition model specifically optimized for the Kurmanji Kurdish dialect

Based on XLS-R architecture

Utilizes Facebook's XLS-R-300M large-scale cross-lingual speech representation model as the foundation

Fine-tuned on Common Voice dataset

Fine-tuned on the Kurmanji dataset from Mozilla Common Voice 7.0

Model Capabilities

Kurmanji Kurdish speech recognition

Speech-to-text

Use Cases

Speech transcription

Kurdish speech transcription

Convert speech content in the Kurmanji dialect to text

Achieved a 26.88% word error rate on the test set

Voice assistants

Kurdish voice interface

Provide voice interaction capabilities for Kurdish users

🚀 wav2vec2-large-xls-r-300m-kurdish

This model is a fine - tuned version of facebook/wav2vec2-xls-r-300m on the MOZILLA - FOUNDATION/COMMON_VOICE_7_0 - KMR dataset. It offers high - performance automatic speech recognition capabilities for the Kurmanji Kurdish language.

🚀 Quick Start

This model is a fine - tuned version of facebook/wav2vec2-xls-r-300m on the MOZILLA - FOUNDATION/COMMON_VOICE_7_0 - KMR dataset. It achieves the following results on the evaluation set:

Loss: 0.2548
Wer: 0.2688

✨ Features

Fine - tuned Model: Based on the pre - trained facebook/wav2vec2-xls-r-300m, fine - tuned on the MOZILLA - FOUNDATION/COMMON_VOICE_7_0 - KMR dataset.
Evaluation Results: Achieves specific loss and WER values on the evaluation set, indicating its performance.

📚 Documentation

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 7e - 05
train_batch_size: 32
eval_batch_size: 1
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon = 1e - 08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 2000
num_epochs: 100.0
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.3161	12.27	2000	0.4199	0.4797
1.0643	24.54	4000	0.2982	0.3721
0.9718	36.81	6000	0.2762	0.3333
0.8772	49.08	8000	0.2586	0.3051
0.8236	61.35	10000	0.2575	0.2865
0.7745	73.62	12000	0.2603	0.2816
0.7297	85.89	14000	0.2539	0.2727
0.7079	98.16	16000	0.2554	0.2681

Framework versions

Transformers 4.16.0.dev0
Pytorch 1.10.1+cu102
Datasets 1.17.1.dev0
Tokenizers 0.11.0

📄 License

This model is released under the Apache - 2.0 license.

📦 Model Index

Property	Details
Model Type	XLS - R - 300M - Kurmanji Kurdish
Training Data	MOZILLA - FOUNDATION/COMMON_VOICE_7_0 - KMR
Results on Evaluation Set	Loss: 0.2548, Wer: 0.2688
Training Hyperparameters	learning_rate: 7e - 05, train_batch_size: 32, etc.
Framework Versions	Transformers 4.16.0.dev0, Pytorch 1.10.1+cu102, etc.

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