Commit
•
2a34599
1
Parent(s):
f2e7e64
Adding the first info to the RAEDME file
Browse files
README.md
CHANGED
|
@@ -1,3 +1,73 @@
|
|
| 1 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
license: cc-by-4.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
+
language:
|
| 3 |
+
- mt
|
| 4 |
+
library_name: nemo
|
| 5 |
+
datasets:
|
| 6 |
+
- common_voice
|
| 7 |
+
thumbnail: null
|
| 8 |
+
tags:
|
| 9 |
+
- automatic-speech-recognition
|
| 10 |
+
- speech
|
| 11 |
+
- audio
|
| 12 |
+
- CTC
|
| 13 |
+
- pytorch
|
| 14 |
+
- NeMo
|
| 15 |
+
- QuartzNet
|
| 16 |
+
- QuartzNet15x5
|
| 17 |
+
- maltese
|
| 18 |
license: cc-by-4.0
|
| 19 |
+
widget:
|
| 20 |
+
model-index:
|
| 21 |
+
- name: stt_mt_quartznet15x5_sp_ep255_64h
|
| 22 |
+
results:
|
| 23 |
+
- task:
|
| 24 |
+
name: Automatic Speech Recognition
|
| 25 |
+
type: automatic-speech-recognition
|
| 26 |
+
dataset:
|
| 27 |
+
name: Common Voice mt
|
| 28 |
+
type: common_voice
|
| 29 |
+
split: test
|
| 30 |
+
args:
|
| 31 |
+
language: mt
|
| 32 |
+
metrics:
|
| 33 |
+
- name: Test WER
|
| 34 |
+
type: wer
|
| 35 |
+
value: <unk>
|
| 36 |
---
|
| 37 |
+
|
| 38 |
+
# stt_mt_quartznet15x5_sp_ep255_64h
|
| 39 |
+
|
| 40 |
+
The "stt_mt_quartznet15x5_sp_ep255_64h" is an acoustic model created with NeMo which is suitable for Automatic Speech Recognition in Maltese.
|
| 41 |
+
|
| 42 |
+
It is the result of fine-tuning the model "QuartzNet15x5Base-En.nemo" [1] with around 64 hours of Maltese data developed by the MASRI Project at the University of Malta between 2019 and 2021. The 64 hours of data were augmented using speed perturbation at speed rates of 0.9 and 1.1. Most of the data is available at the the MASRI Project homepage https://www.um.edu.mt/projects/masri/.
|
| 43 |
+
|
| 44 |
+
The specific list of corpora used to fine-tune the model is:
|
| 45 |
+
|
| 46 |
+
- MASRI-HEADSET v2 (6h39m)
|
| 47 |
+
- MASRI-Farfield (9h37m)
|
| 48 |
+
- MASRI-Booths (2h27m)
|
| 49 |
+
- MASRI-MEP (1h17m)
|
| 50 |
+
- MASRI-COMVO (7h29m)
|
| 51 |
+
- MASRI-TUBE (13h17m)
|
| 52 |
+
- MASRI-MERLIN (25h18m) *Not available at the MASRI Project homepage
|
| 53 |
+
|
| 54 |
+
The fine-tuning process was perform during October (2022) in the servers of the Language and Voice Lab (https://lvl.ru.is/) at Reykjavík University (Iceland) by Carlos Daniel Hernández Mena.
|
| 55 |
+
|
| 56 |
+
```bibtex
|
| 57 |
+
@misc{mena2022quartznet15x5maltese,
|
| 58 |
+
title={Acoustic Model in Maltese: stt_mt_quartznet15x5_sp_ep255_64h.},
|
| 59 |
+
author={Hernandez Mena, Carlos Daniel},
|
| 60 |
+
year={2022},
|
| 61 |
+
url={https://huggingface.co/carlosdanielhernandezmena/stt_mt_quartznet15x5_sp_ep255_64h},
|
| 62 |
+
}
|
| 63 |
+
```
|
| 64 |
+
|
| 65 |
+
# Acknowledgements
|
| 66 |
+
|
| 67 |
+
The MASRI Project is funded by the University of Malta Research Fund Awards. We want to thank to Merlin Publishers (Malta) for provinding the audiobooks used to create the MASRI-MERLIN Corpus.
|
| 68 |
+
|
| 69 |
+
Special thanks to Jón Guðnason, head of the Language and Voice Lab for providing computational power to make this model possible. We also want to thank to the "Language Technology Programme for Icelandic 2019-2023" which is managed and coordinated by Almannarómur, and it is funded by the Icelandic Ministry of Education, Science and Culture.
|
| 70 |
+
|
| 71 |
+
# References
|
| 72 |
+
|
| 73 |
+
[1] For Web Download: https://catalog.ngc.nvidia.com/orgs/nvidia/models/nemospeechmodels/files
|