wav2vec2-large-voxrex-npsc-bokmaal Open-source Speech Recognition Model - Accurately Recognize Norwegian Written Language (Bokmål)

Wav2vec2 Large Voxrex Npsc Bokmaal

Developed by NbAiLab

This is a large automatic speech recognition (ASR) model based on the wav2vec2 architecture, specifically trained for Norwegian written language (Bokmål).

Speech Recognition

Transformers

Open Source License:Apache-2.0 #Norwegian speech recognition #Low word error rate #Bokmål dialect

Downloads 35

Release Time : 3/2/2022

Model Overview

The model is trained using the NbAiLab/NPSC dataset and can convert Norwegian written language (Bokmål) speech into text.

Model Features

High-precision Norwegian recognition

Achieved a word error rate of 7.03% and a character error rate of 2.69% on the NPSC test set.

Domain-specific optimization

Specifically trained and optimized for Norwegian written language (Bokmål).

Based on wav2vec2 architecture

Utilizes the advanced wav2vec2-large-voxrex architecture with powerful speech feature extraction capabilities.

Model Capabilities

Norwegian speech recognition

Speech-to-text

Automatic speech transcription

Use Cases

Speech transcription

Norwegian meeting minutes

Automatically transcribe Norwegian meeting recordings into text records

Highly accurate transcription results

Voice assistants

Provide speech recognition capabilities for Norwegian voice assistants

Accessibility services

Hearing impairment assistance

Provide real-time speech-to-text services for people with hearing impairments

🚀 wav2vec2-large-voxrex-npsc-bokmaal

This model, wav2vec2-large-voxrex-npsc-bokmaal, is designed for automatic speech recognition. It was trained from scratch on the None dataset and achieves notable results on the evaluation set, offering a reliable solution for speech - related tasks.

🚀 Quick Start

This section will be updated with quick - start instructions once they are available.

✨ Features

High - Performance Speech Recognition: Achieves low WER and CER on the NPSC dataset, demonstrating its effectiveness in automatic speech recognition.
Customizable Training: Trained with specific hyperparameters that can be adjusted according to different requirements.

📦 Installation

Installation instructions will be provided in the future.

📚 Documentation

Model Information

Property	Details
Model Type	wav2vec2-large-voxrex-npsc-bokmaal
Training Data	None
License	Apache - 2.0
Tags	generated_from_trainer, automatic - speech - recognition, NbAiLab/NPSC, robust - speech - event, false, nb - NO, hf - asr - leaderboard

Evaluation Results

This model achieves the following results on the evaluation set:

Loss: 0.1311
Wer: 0.1038

On the NPSC dataset (16K_mp3_bokmaal):

Metric	Value
Test (Bokmål) WER	0.07028972259374369
Test (Bokmål) CER	0.026870600821650645

Model description

More information about the model description will be provided later.

Intended uses & limitations

Details about the intended uses and limitations will be added in the future.

Training and evaluation data

Information regarding the training and evaluation data will be updated soon.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 8.379967082059723e - 06
train_batch_size: 16
eval_batch_size: 16
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 32
optimizer: Adam with betas=(0.9,0.999) and epsilon = 1e - 08
lr_scheduler_type: linear
num_epochs: 0.1
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.2127	0.32	500	0.1335	0.1047
0.1976	0.64	1000	0.1309	0.1039
0.1887	0.97	1500	0.1306	0.1040
0.18	1.29	2000	0.1311	0.1038

Framework versions

Transformers 4.17.0.dev0
Pytorch 1.10.2+cu113
Datasets 1.18.4.dev0
Tokenizers 0.11.0

📄 License

This model is released 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