update README
Browse files
README.md
CHANGED
|
@@ -6,7 +6,7 @@ FunASR SenseVoice on Axera, official repo: https://github.com/FunAudioLLM/SenseV
|
|
| 6 |
|
| 7 |
## 功能
|
| 8 |
- 语音识别
|
| 9 |
-
- 自动识别语言
|
| 10 |
- 情感识别
|
| 11 |
- 自动标点
|
| 12 |
|
|
@@ -47,6 +47,19 @@ RTF: 0.03026517820946964 Latency: 0.15689468383789062s Total length: 5.184s
|
|
| 47 |
['呢几个字。', '都表达唔到，我想讲嘅意。', '思。']
|
| 48 |
```
|
| 49 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 50 |
## 技术讨论
|
| 51 |
|
| 52 |
- Github issues
|
|
|
|
| 6 |
|
| 7 |
## 功能
|
| 8 |
- 语音识别
|
| 9 |
+
- 自动识别语言(支持中文、英文、粤语、日语、韩语)
|
| 10 |
- 情感识别
|
| 11 |
- 自动标点
|
| 12 |
|
|
|
|
| 47 |
['呢几个字。', '都表达唔到，我想讲嘅意。', '思。']
|
| 48 |
```
|
| 49 |
|
| 50 |
+
## 准确率
|
| 51 |
+
|
| 52 |
+
使用WER(Word-Error-Rate)作为评价标准
|
| 53 |
+
|
| 54 |
+
**WER = 0.0389**
|
| 55 |
+
|
| 56 |
+
### 复现测试结果
|
| 57 |
+
|
| 58 |
+
```
|
| 59 |
+
./download_datasets.sh
|
| 60 |
+
python test_wer.py -d datasets -l zh
|
| 61 |
+
```
|
| 62 |
+
|
| 63 |
## 技术讨论
|
| 64 |
|
| 65 |
- Github issues
|