Update README.md
Browse files
README.md
CHANGED
|
@@ -31,8 +31,8 @@ This version is trained on a significantly larger dataset compared to **[V1](htt
|
|
| 31 |
## 🦻 Why V2 has higher WER than V1?
|
| 32 |
V2 shows a higher Word Error Rate (40.5%) compared to V1 (31.5%) mainly because of the evaluation dataset.
|
| 33 |
V2 was evaluated on a much larger and significantly harder test set, which includes strong regional dialects such as Hewlêr and Mukriyanî.
|
| 34 |
-
These dialects contain heavier pronunciation variations that are more challenging for ASR models.
|
| 35 |
-
Real-World Performance
|
| 36 |
Despite the higher validation WER, V2 performs very well on real-world audio.<br/>
|
| 37 |
**_In manual testing across different audio scenarios:_**<br/>
|
| 38 |
When speakers used clear and standard Sorani (academic pronunciation), the model produced correct Kurdish text in every single test.
|
|
|
|
| 31 |
## 🦻 Why V2 has higher WER than V1?
|
| 32 |
V2 shows a higher Word Error Rate (40.5%) compared to V1 (31.5%) mainly because of the evaluation dataset.
|
| 33 |
V2 was evaluated on a much larger and significantly harder test set, which includes strong regional dialects such as Hewlêr and Mukriyanî.
|
| 34 |
+
These dialects contain heavier pronunciation variations that are more challenging for ASR models.<br />
|
| 35 |
+
**Real-World Performance**<br/>
|
| 36 |
Despite the higher validation WER, V2 performs very well on real-world audio.<br/>
|
| 37 |
**_In manual testing across different audio scenarios:_**<br/>
|
| 38 |
When speakers used clear and standard Sorani (academic pronunciation), the model produced correct Kurdish text in every single test.
|