File size: 2,456 Bytes
633a9e7 54c4dba 633a9e7 b722a56 633a9e7 96f8bab f471100 b722a56 d118fdd 633a9e7 b722a56 633a9e7 54c4dba 633a9e7 54c4dba 99a28ed cce5842 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 |
---
language: hr
datasets:
- parlaspeech-hr
tags:
- audio
- automatic-speech-recognition
- parlaspeech
widget:
- example_title: example 1
src: https://huggingface.co/classla/wav2vec2-xls-r-sabor-hr/raw/main/00020578b.flac.wav
- example_title: example 2
src: https://huggingface.co/classla/wav2vec2-xls-r-sabor-hr/raw/main/00020570a.flac.wav
---
# wav2vec2-xls-r-parlaspeech-hr
This model is based on the [facebook/wav2vec2-xls-r-300m model](https://huggingface.co/facebook/wav2vec2-xls-r-300m) and was fine-tuned with 72 hours of recordings and transcripts from the Croatian parliament. This training dataset is an early result of the second iteration of the [ParlaMint project](https://www.clarin.eu/content/parlamint-towards-comparable-parliamentary-corpora) inside which the dataset will be extended and published under the name ParlaSpeech-HR and an open licence.
The efforts resulting in this model were coordinated by Nikola Ljubešić, the rough manual data alignment was performed by Ivo-Pavao Jazbec, the method for fine automatic data alignment from [Plüss et al.](https://arxiv.org/abs/2010.02810) was applied by Vuk Batanović and Lenka Bajčetić, while the final modelling was performed by Peter Rupnik.
Initial evaluation on partially noisy data showed the model to achieve a word error rate of 13.68% and a character error rate of 4.56%.
## Usage in `transformers`
```python
from transformers import Wav2Vec2Processor, Wav2Vec2ForCTC
from datasets import Audio
import soundfile as sf
import torch
import os
# load model and tokenizer
processor = Wav2Vec2Processor.from_pretrained(
"classla/wav2vec2-xls-r-sabor-hr")
model = Wav2Vec2ForCTC.from_pretrained("classla/wav2vec2-xls-r-sabor-hr")
# download the example wav files:
os.system("curl https://huggingface.co/classla/wav2vec2-xls-r-sabor-hr/raw/main/00020570a.flac.wav")
# read the wav file as datasets.Audio object
audio = Audio(sampling_rate=16000).decode_example("00020570a.flac.wav")
# remove the raw wav file
os.system("rm 00020570a.flac.wav")
# tokenize
input_values = processor(
audio["array"], return_tensors="pt", padding=True,
sampling_rate=16000).input_values
# retrieve logits
logits = model(input_values).logits
# take argmax and decode
predicted_ids = torch.argmax(logits, dim=-1)
transcription = processor.batch_decode(predicted_ids)
# transcription: ['veliki broj poslovnih subjekata posluje sa minusom velik dio']
```
|