Datasets:
add note about newlines
Browse files
README.md
CHANGED
|
@@ -151,6 +151,8 @@ size_categories:
|
|
| 151 |
We release all evaluation data and scripts for further analysis and reproduction of the accompanying paper: [A comparison of translation performance between DeepL and Supertext](https://arxiv.org/abs/2502.02577).
|
| 152 |
The data consists of document-level translations by Supertext and DeepL as well as accompanying ratings by professional translators. Please find more details in the paper.
|
| 153 |
|
|
|
|
|
|
|
| 154 |
``` python
|
| 155 |
# for each language pair, there is a separate subset
|
| 156 |
data = load_dataset("Supertext/mt-doclevel-ab-test", "en-deCH")
|
|
|
|
| 151 |
We release all evaluation data and scripts for further analysis and reproduction of the accompanying paper: [A comparison of translation performance between DeepL and Supertext](https://arxiv.org/abs/2502.02577).
|
| 152 |
The data consists of document-level translations by Supertext and DeepL as well as accompanying ratings by professional translators. Please find more details in the paper.
|
| 153 |
|
| 154 |
+
Please note that the empty lines correspond to paragraph boundaries (i.e., double line breaks) in the original documents.
|
| 155 |
+
|
| 156 |
``` python
|
| 157 |
# for each language pair, there is a separate subset
|
| 158 |
data = load_dataset("Supertext/mt-doclevel-ab-test", "en-deCH")
|