wav2vec2-large-voxrex-swedish

Wav2vec 2.0 large VoxRex Swedish (C)

Finetuned version of KBs VoxRex large model using Swedish radio broadcasts, NST and Common Voice data. Evalutation without a language model gives the following: 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.

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: Paper is is here.

Performance*

<center><del>*<i>Chart shows performance without the additional 20k steps of Common Voice fine-tuning</i></del></center>

Training

This model has been fine-tuned for 120000 updates on NST + CommonVoice<del> 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]</del>.

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])

Citation

https://arxiv.org/abs/2205.03026

@inproceedings{malmsten2022hearing,
  title={Hearing voices at the national library : a speech corpus and acoustic model for the Swedish language},
  author={Malmsten, Martin and Haffenden, Chris and B{\"o}rjeson, Love},
  booktitle={Proceeding of Fonetik 2022 : Speech, Music and Hearing Quarterly Progress and Status Report, TMH-QPSR},
  volume={3},
  year={2022}
}