Commit
•
a4300f7
1
Parent(s):
6910d55
Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,23 @@
|
|
|
|
|
| 1 |
---
|
|
|
|
|
|
|
| 2 |
license: apache-2.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
```
|
| 2 |
---
|
| 3 |
+
language:
|
| 4 |
+
- ch
|
| 5 |
license: apache-2.0
|
| 6 |
+
base_model: openai/whisper-small
|
| 7 |
+
datasets:
|
| 8 |
+
- TingChen-ppmc/Shanghai_Dialect_Conversational_Speech_Corpus
|
| 9 |
+
metrics:
|
| 10 |
+
- wer
|
| 11 |
+
- cer
|
| 12 |
---
|
| 13 |
+
```
|
| 14 |
+
|
| 15 |
+
This is a finetuned version of [Whisper-small](openai/whisper-small) on [Shanghai_Dialect_Conversational_Speech_Corpus](https://huggingface.co/datasets/TingChen-ppmc/Shanghai_Dialect_Conversational_Speech_Corpus)
|
| 16 |
+
|
| 17 |
+
# Results
|
| 18 |
+
|
| 19 |
+
CER: 56.07
|
| 20 |
+
|
| 21 |
+
WER: 75.62
|
| 22 |
+
|
| 23 |
+
Chinese Word Error Rate (WER) is based on [Jieba](https://github.com/fxsjy/jieba) Word segmentation
|