Wav2vec2 Large Voxrex Swedish

🚀 Wav2vec 2.0 large VoxRex Swedish (C)

This is a fine - tuned version of KBs VoxRex large model. It uses Swedish radio broadcasts, NST and Common Voice data to enhance the performance of automatic speech recognition for the Swedish language.

🚀 Quick Start

When using this model, make sure that your speech input is sampled at 16kHz.

Update 2022 - 01 - 10: Updated to VoxRex - C version. Update 2022 - 05 - 16: The related paper is here.

✨ Features

• Finetuned based on KBs VoxRex large model.
• Utilizes Swedish radio broadcasts, NST, and Common Voice data.
• Achieves low Word Error Rate (WER) in speech recognition tasks for the Swedish language.

📦 Installation

No specific installation steps are provided in the original document, so this section is skipped.

💻 Usage Examples

Basic Usage

The model can be used directly (without a language model) as follows:

import torch
import torchaudio
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
test_dataset = load_dataset("common_voice", "sv-SE", split="test[:2%]")
processor = Wav2Vec2Processor.from_pretrained("KBLab/wav2vec2-large-voxrex-swedish")
model = Wav2Vec2ForCTC.from_pretrained("KBLab/wav2vec2-large-voxrex-swedish")
resampler = torchaudio.transforms.Resample(48_000, 16_000)
# Preprocessing the datasets.
# We need to read the aduio files as arrays
def speech_file_to_array_fn(batch):
    speech_array, sampling_rate = torchaudio.load(batch["path"])
    batch["speech"] = resampler(speech_array).squeeze().numpy()
    return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"][:2], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
    logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
predicted_ids = torch.argmax(logits, dim=-1)
print("Prediction:", processor.batch_decode(predicted_ids))
print("Reference:", test_dataset["sentence"][:2])

📚 Documentation

Performance

Evalutation without a language model gives the following results:

• WER for NST + Common Voice test set (2% of total sentences) is 2.5%.
• WER for Common Voice test set is 8.49% directly and 7.37% with a 4 - gram language model.

Training

This model has been fine - tuned for 120000 updates on NST + CommonVoice and then for an additional 20000 updates on CommonVoice only. The additional fine - tuning on CommonVoice hurts performance on the NST+CommonVoice test set somewhat and, unsurprisingly, improves it on the CommonVoice test set. It seems to perform generally better though [citation needed].

🔧 Technical Details

No specific technical details meeting the requirement (>50 words) are provided in the original document, so this section is skipped.

📄 License

This model is licensed under cc0 - 1.0.

📋 Information Table

📖 Citation

https://arxiv.org/abs/2205.03026

@misc{malmsten2022hearing,
      title={Hearing voices at the National Library -- a speech corpus and acoustic model for the Swedish language}, 
      author={Martin Malmsten and Chris Haffenden and Love Börjeson},
      year={2022},
      eprint={2205.03026},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}