---
language: sv
datasets:
- common_voice
- NST Swedish ASR Database
- P4
metrics:
- wer
tags:
- audio
- automatic-speech-recognition
- speech
license: cc0-1.0
model-index:
- name: Wav2vec 2.0 large VoxRex Swedish 4-gram
---

# KBLab's wav2vec 2.0 large VoxRex Swedish (C) with 4-gram model
Training of the acoustic model is the work of KBLab. See [VoxRex-C](https://huggingface.co/KBLab/wav2vec2-large-voxrex-swedish) for more details. This repo extends the acoustic model with a social media 4-gram language model for boosted performance.

## Model description
VoxRex-C is extended with a 4-gram language model estimated from a subset extracted from [The Swedish Culturomics Gigaword Corpus](https://spraakbanken.gu.se/resurser/gigaword) from Språkbanken. The subset contains 40M words from the social media genre between 2010 and 2015.

## How to use
Audio should be downsampled to 16kHz.

```python
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])
```

## Training procedure
Text data for the n-gram model is pre-processed by removing characters not part of the wav2vec 2.0 vocabulary and uppercasing all characters. After pre-processing and storing each text sample on a new line in a text file, a [KenLM](https://github.com/kpu/kenlm) model is estimated. See [this tutorial](https://huggingface.co/blog/wav2vec2-with-ngram) for more details.

## Evaluation results
The model was evaluated on the full Common Voice test set version 6.1. VoxRex-C achieved a WER of 9.03% without the language model and 6.47% with the language model.