add <LunYu> and <ZhongYong> dataset
This commit is contained in:
1 parent
cf9b659262
commit
8a60d1478e
5 files changed
+3542
-7
No files matched your search
+4
-2
@@ -9,6 +9,8 @@ A classical Chinese translation corpus dataset.(Under construction)
|
||||
Translated data of classical Chinese:
|
||||
|
||||
- Historical Records, by Sima Qian (29 for now)
|
||||
- LunYu (20 articles)
|
||||
- ZhongYong (33 chapters)
|
||||
|
||||
Data format
|
||||
|
||||
@@ -34,8 +36,8 @@ Statement
|
||||
|
||||
| |Samples|Max Length|Min Length|Average Length|Median Length|Standard Deviation|
|
||||
|----|:---:|:------:|:-----:|:-------:|:-------:|:-------:|
|
||||
|source|7026 |111 |1 |16.02 |15.0 |11.08 |
|
||||
|target|7026 |202 |2 |25.99 |22.0 |18.52 |
|
||||
|source|7841 |180 |1 |17.67 |15.0 |13.54 |
|
||||
|target|7841 |280 |2 |29.20 |24.0 |23.28 |
|
||||
|
||||
## Contribution
|
||||
|
||||
|
||||
Reference in new issue
Block a user