wac2vec-lllfantomlll Open-source Speech Recognition Model - Low Word Error Rate for Precise Speech Recognition

Wac2vec Lllfantomlll

Developed by lllFaNToMlll

A speech recognition model fine-tuned based on facebook/wav2vec2-base, achieving a word error rate of 0.3417 on the evaluation set.

Speech Recognition

Transformers

Open Source License:Apache-2.0 #Speech Recognition #Low Word Error Rate #wav2vec2 Fine-tuning

Downloads 27

Release Time : 6/11/2022

Model Overview

This model is a fine-tuned version of the wav2vec2 model for speech recognition, capable of converting speech to text.

Model Features

Low Word Error Rate

Achieved a word error rate of 0.3417 on the evaluation set, demonstrating good performance.

Based on wav2vec2 Architecture

Uses facebook's wav2vec2-base as the base model, featuring excellent speech feature extraction capabilities.

Fine-tuned

After 30 training epochs with gradually adjusted learning rates, the model's performance steadily improved.

Model Capabilities

Speech Recognition

Speech to Text

Use Cases

Speech Transcription

Meeting Minutes

Automatically convert meeting recordings into text transcripts

Accuracy approximately 65.83% (word error rate 0.3417)

Voice Notes

Convert voice notes into searchable text

🚀 wac2vec-lllfantomlll

This model is a fine - tuned version of [facebook/wav2vec2 - base](https://huggingface.co/facebook/wav2vec2 - base), aiming to provide better performance in relevant speech - related tasks.

🚀 Quick Start

This model is a fine - tuned version of [facebook/wav2vec2 - base](https://huggingface.co/facebook/wav2vec2 - base) on the None dataset. It achieves the following results on the evaluation set:

Loss: 0.5560
Wer: 0.3417

📚 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: 0.0001
train_batch_size: 8
eval_batch_size: 8
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon = 1e - 08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 1000
num_epochs: 30
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.5768	1.0	500	2.0283	1.0238
0.9219	2.01	1000	0.5103	0.5022
0.4497	3.01	1500	0.4746	0.4669
0.3163	4.02	2000	0.4144	0.4229
0.2374	5.02	2500	0.4186	0.4161
0.2033	6.02	3000	0.4115	0.3975
0.1603	7.03	3500	0.4424	0.3817
0.1455	8.03	4000	0.4151	0.3918
0.1276	9.04	4500	0.4940	0.3798
0.108	10.04	5000	0.4580	0.3688
0.1053	11.04	5500	0.4243	0.3700
0.0929	12.05	6000	0.4999	0.3727
0.0896	13.05	6500	0.4991	0.3624
0.0748	14.06	7000	0.4924	0.3602
0.0681	15.06	7500	0.4908	0.3544
0.0619	16.06	8000	0.5021	0.3559
0.0569	17.07	8500	0.5448	0.3518
0.0549	18.07	9000	0.4919	0.3508
0.0478	19.08	9500	0.4704	0.3513
0.0437	20.08	10000	0.5058	0.3555
0.0421	21.08	10500	0.5127	0.3489
0.0362	22.09	11000	0.5439	0.3527
0.0322	23.09	11500	0.5418	0.3469
0.0327	24.1	12000	0.5298	0.3422
0.0292	25.1	12500	0.5511	0.3426
0.0246	26.1	13000	0.5349	0.3472
0.0251	27.11	13500	0.5646	0.3391
0.0214	28.11	14000	0.5821	0.3424
0.0217	29.12	14500	0.5560	0.3417

Framework versions

Transformers 4.17.0
Pytorch 1.11.0+cu113
Datasets 1.18.3
Tokenizers 0.12.1

📄 License

This model 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