File size: 1,847 Bytes
2e11669 c869ad3 2e11669 d9e8a10 2e11669 d9e8a10 2e11669 a51f5db e676007 a51f5db 2e11669 |
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 65 66 |
---
language: "rw"
thumbnail:
pipeline_tag: automatic-speech-recognition
tags:
- Coqui
- Deepspeech
- LSTM
license: "apache-2.0"
datasets:
- commonvoice
metrics:
- wer
---
**Model card - Kinyarwanda coqui STT model**
**Model details**
- Kinyarwanda Speech to text model
- Developed by [Digital Umuganda](digitalumuganda.com)
- Model based from: Baidu Deepspeech end to end RNN model
- paper: [deepspeech end to end STT](https://arxiv.org/pdf/1412.5567.pdf)
- Documentation on model: [deepspeech documentation](https://deepspeech.readthedocs.io/)
- License: Mozilla 2.0 License
- Feedback on the model: samuel@digitalumuganda.com
**Intended use cases**
- Intended to be used for
- simple keyword spotting
- simple transcribing
- transfer learning for better kinyarwanda and african language models
- Intended to be used by:
- App developpers
- various organizations who want to transcribe kinyarwanda recordings
- ML researchers
- other researchers in Kinyarwanda and tech usage in kinyarwanda (e.g. Linguists, journalists)
- Not intended to be used as:
- a fully fledged voice assistant
- voice recognition application
- Multiple languages STT
- language detection
**Factors**
- Anti-bias: these are bias that can influence the accuracy of the model
- Gender
- accents and dialects
- age
- Voice quality: factors that can influence the accuracy of the model
- Background noise
- short sentences
- Voice format: voices must be converted to the wav format
- wav format
**Metrics**
- word error rate on the Common Voice Kinyarwanda test set
|Test Corpus|WER|
|-----------|---|
|Common Voice|39.1\%|
**Training data**
- [common voice crowdsource website](https://commonvoice.mozilla.org/en/datasets)
**Evaluation data**
- [common voice crowdsource website](https://commonvoice.mozilla.org/en/datasets)
|