language:
- zgh
- kab
- shi
- rif
- tzm
- shy
license: cc-by-4.0
library_name: nemo
datasets:
- mozilla-foundation/common_voice_18_0
thumbnail: null
tags:
- automatic-speech-recognition
- speech
- audio
- CTC
- FastConformer
- Transformer
- NeMo
- pytorch
model-index:
- name: stt_zgh_fastconformer_ctc_small
results:
- task:
type: Automatic Speech Recognition
name: automatic-speech-recognition
dataset:
name: Mozilla Common Voice 18.0
type: mozilla-foundation/common_voice_18_0
config: zgh
split: test
args:
language: zgh
metrics:
- name: Test WER
type: wer
value: 64.17
- task:
type: Automatic Speech Recognition
name: automatic-speech-recognition
dataset:
name: Mozilla Common Voice 18.0
type: mozilla-foundation/common_voice_18_0
config: zgh
split: test
args:
language: zgh
metrics:
- name: Test CER
type: cer
value: 21.54
- task:
type: Automatic Speech Recognition
name: automatic-speech-recognition
dataset:
name: Mozilla Common Voice 18.0
type: mozilla-foundation/common_voice_18_0
config: kab
split: test
args:
language: kab
metrics:
- name: Test WER
type: wer
value: 34.87
- task:
type: Automatic Speech Recognition
name: automatic-speech-recognition
dataset:
name: Mozilla Common Voice 18.0
type: mozilla-foundation/common_voice_18_0
config: kab
split: test
args:
language: kab
metrics:
- name: Test CER
type: cer
value: 13.11
metrics:
- wer
- cer
pipeline_tag: automatic-speech-recognition
Model Overview
NVIDIA NeMo: Training
To train, fine-tune or play with the model you will need to install NVIDIA NeMo. We recommend you install it after you've installed latest Pytorch version.
pip install nemo_toolkit['asr']
How to Use this Model
The model is available for use in the NeMo toolkit [3], and can be used as a pre-trained checkpoint for inference or for fine-tuning on another dataset.
Automatically instantiate the model
import nemo.collections.asr as nemo_asr
asr_model = nemo_asr.models.ASRModel.from_pretrained("ayymen/stt_zgh_fastconformer_ctc_small")
Transcribing using Python
First, let's get a sample
wget https://dldata-public.s3.us-east-2.amazonaws.com/2086-149220-0033.wav
Then simply do:
asr_model.transcribe(['2086-149220-0033.wav'])
Transcribing many audio files
python [NEMO_GIT_FOLDER]/examples/asr/transcribe_speech.py pretrained_name="ayymen/stt_zgh_fastconformer_ctc_small" audio_dir="<DIRECTORY CONTAINING AUDIO FILES>"
Input
This model accepts 16000 KHz Mono-channel Audio (wav files) as input.
Output
This model provides transcribed speech as a string for a given audio sample.
Model Architecture
Training
The model was fine-tuned from an older checkpoint on a NVIDIA GeForce RTX 4050 Laptop GPU.
Datasets
Common Voice 18 kab and zgh splits, Tatoeba (kab, ber, shy), and bible readings in Tachelhit and Tarifit.
Performance
Metrics are computed on the cleaned, non-punctuated test sets of zgh and kab (converted to Tifinagh).
Limitations
Eg: Since this model was trained on publicly available speech datasets, the performance of this model might degrade for speech which includes technical terms, or vernacular that the model has not been trained on. The model might also perform worse for accented speech.