Update README.md
Browse files
README.md
CHANGED
|
@@ -11,4 +11,27 @@ datasets:
|
|
| 11 |
widget:
|
| 12 |
- text: "Muraho neza, murakaza neza mu Rwanda."
|
| 13 |
example_title: "Muraho neza, murakaza neza mu Rwanda."
|
| 14 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 11 |
widget:
|
| 12 |
- text: "Muraho neza, murakaza neza mu Rwanda."
|
| 13 |
example_title: "Muraho neza, murakaza neza mu Rwanda."
|
| 14 |
+
---
|
| 15 |
+
|
| 16 |
+
**Model card - Kinyarwanda TTS model**
|
| 17 |
+
|
| 18 |
+
**Model details**
|
| 19 |
+
- Kinyarwanda Text to Speech model
|
| 20 |
+
- Developed by [Digital Umuganda](digitalumuganda.com), [Arxia](https://www.arxia.com/home.html) and [Zevo Tech](https://zevo-tech.com/)
|
| 21 |
+
- Model based from: Fastspeech and Waveglow
|
| 22 |
+
- License: Mozilla 2.0 License
|
| 23 |
+
- Feedback on the model: [email protected]
|
| 24 |
+
|
| 25 |
+
**Metrics**
|
| 26 |
+
- We use Mean Opinion Score (MOS) to evaluate the model with a maximum score being 5
|
| 27 |
+
|
| 28 |
+
|Test Corpus|MOS|
|
| 29 |
+
|-----------|---|
|
| 30 |
+
|Custom phrases|3|
|
| 31 |
+
|
| 32 |
+
**Challenges**
|
| 33 |
+
- The model does not always capture the Kinyarwanda tones
|
| 34 |
+
|
| 35 |
+
**Recommendations**
|
| 36 |
+
- Use a tonal dictionary to train future models
|
| 37 |
+
- Add numbers and symbols Dictionary
|