Datasets:
Tasks:
Sentence Similarity
Formats:
csv
Languages:
German
Size:
10M - 100M
ArXiv:
Tags:
sentence-transformers
License:
Update README.md
Browse files
README.md
CHANGED
@@ -29,12 +29,24 @@ The English texts were machine translated back into German. This is how the para
|
|
29 |
| **sum** |. **21,292,789** |
|
30 |
|
31 |
## To-do
|
32 |
-
- add info
|
33 |
-
- add info about exact size (size per corpus)
|
34 |
-
- add copyright info and citations of individual datasets
|
35 |
- add column description
|
36 |
- upload dataset
|
37 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
38 |
## Citations & Acknowledgements
|
39 |
|
40 |
**OpenSubtitles**\
|
|
|
29 |
| **sum** |. **21,292,789** |
|
30 |
|
31 |
## To-do
|
32 |
+
- add copyright info of individual datasets
|
|
|
|
|
33 |
- add column description
|
34 |
- upload dataset
|
35 |
|
36 |
+
## Back translation
|
37 |
+
We have made the back translation from English to German with the help of [Fairseq](https://github.com/facebookresearch/fairseq).
|
38 |
+
We used the `transformer.wmt19.en-de` model for this purpose:
|
39 |
+
|
40 |
+
```python
|
41 |
+
en2de = torch.hub.load(
|
42 |
+
"pytorch/fairseq",
|
43 |
+
"transformer.wmt19.en-de",
|
44 |
+
checkpoint_file="model1.pt:model2.pt:model3.pt:model4.pt",
|
45 |
+
tokenizer="moses",
|
46 |
+
bpe="fastbpe",
|
47 |
+
)
|
48 |
+
```
|
49 |
+
|
50 |
## Citations & Acknowledgements
|
51 |
|
52 |
**OpenSubtitles**\
|