Wav2vec2 Xlsr 1b Finnish Lm V2

🚀 Wav2Vec2 XLS-R for Finnish ASR

This acoustic model is designed for Finnish Automatic Speech Recognition (ASR). It's a fine - tuned version of facebook/wav2vec2-xls-r-1b, trained with 275.6 hours of Finnish transcribed speech data. Wav2Vec2 XLS - R was introduced in this paper and first released at this page. This repository also includes a Finnish KenLM language model for use in the decoding phase with the acoustic model.

🚀 Quick Start

To use this model for Finnish ASR, check the run-finnish-asr-models.ipynb notebook in this repository for a detailed example.

✨ Features

• Fine - tuned for Finnish: Specifically tailored for Finnish ASR, leveraging 275.6 hours of Finnish transcribed speech data.
• Multilingual Pretrained Base: Built on the large - scale multilingual pretrained model Wav2Vec2 XLS - R, pretrained on 436k hours of unlabeled speech across 128 languages.
• Language Model Included: Comes with a Finnish KenLM language model for decoding.

📦 Installation

No specific installation steps are provided in the original document.

💻 Usage Examples

No code examples are provided in the original document.

📚 Documentation

Model description

Wav2Vec2 XLS - R is a large - scale multilingual pretrained speech model from Facebook AI. It's pretrained on 436k hours of unlabeled speech from various sources like VoxPopuli, MLS, CommonVoice, BABEL, and VoxLingua107, using the wav2vec 2.0 objective in 128 languages. You can learn more about the pretrained model from this blog and this paper. This model is a fine - tuned version of the 1 - billion - parameter variant for Finnish ASR.

Intended uses & limitations

• Intended use: This model can be used for Finnish ASR (speech - to - text) tasks.
• Limitations:
  • Audio length: The model was fine - tuned with audio samples of maximum 20 seconds, so it may work best for short audios of similar length. For very long audio files, you can try the audio chunking method introduced in this blog post if you encounter out - of - memory errors.
  • Domain generalization: A large part of the fine - tuning data is from the Finnish Parliament dataset, so the model may not generalize well to different domains like daily spoken Finnish with dialects.
  • Gender bias: The datasets are mostly dominated by adult male voices, so the model may not perform as well for children and women's speeches.
  • Language model generalization: The Finnish KenLM language model used in decoding was trained with text from audio transcriptions and a subset of Finnish Wikipedia. It may not generalize well to different language styles, such as daily spoken language with dialects. It may be beneficial to train your own KenLM language model for your specific domain.

Training data

This model was fine - tuned with 275.6 hours of Finnish transcribed speech data from the following datasets:

The datasets were filtered to include audio samples of maximum 20 seconds in length.

Training procedure

• Event: This model was trained during the Robust Speech Challenge Event organized by Hugging Face.
• Hardware: Training was done on a Tesla V100 GPU, sponsored by OVHcloud.
• Training script: The training script was provided by Hugging Face and is available here. Only the data loading was modified for custom datasets.
• KenLM language model training: The training of the 5 - gram KenLM language model followed the blog post tutorial provided by Hugging Face. The training data included text transcriptions of the audio training data and 100k random samples of the cleaned Finnish Wikipedia (August 2021) dataset.

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e - 05
• train_batch_size: 32
• eval_batch_size: 8
• seed: 42
• optimizer: 8 - bit Adam with betas=(0.9, 0.999) and epsilon = 1e - 08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 10
• mixed_precision_training: Native AMP

The pretrained facebook/wav2vec2-xls-r-1b model was initialized with the following hyperparameters:

• attention_dropout: 0.094
• hidden_dropout: 0.047
• feat_proj_dropout: 0.04
• mask_time_prob: 0.082
• layerdrop: 0.041
• activation_dropout: 0.055
• ctc_loss_reduction: "mean"

Training results

Framework versions

• Transformers 4.17.0.dev0
• Pytorch 1.10.2+cu102
• Datasets 1.18.3
• Tokenizers 0.11.0

Evaluation results

Evaluation was done with the Common Voice 7.0 Finnish test split. To evaluate this model, run the eval.py script in this repository:

python3 eval.py --model_id aapot/wav2vec2-xlsr-1b-finnish-lm-v2 --dataset mozilla-foundation/common_voice_7_0 --config fi --split test

This model (the first row of the table) achieves the following WER (Word Error Rate) and CER (Character Error Rate) results compared to other models:

🔧 Technical Details

The model is based on the Wav2Vec2 XLS - R architecture. The fine - tuning process involves adjusting the pretrained model's parameters using Finnish transcribed speech data. The use of the KenLM language model in the decoding phase helps improve the accuracy of speech recognition.

📄 License

This model is released under the Apache - 2.0 license.

Team Members

• Aapo Tanskanen, Hugging Face profile, LinkedIn profile
• Rasmus Toivanen, Hugging Face profile, LinkedIn profile

Feel free to contact us for more details 🤗