Mainak Manna
commited on
Commit
•
c756558
1
Parent(s):
c2fe29a
First version of the model
Browse files
README.md
ADDED
@@ -0,0 +1,67 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
|
2 |
+
---
|
3 |
+
language: French English
|
4 |
+
tags:
|
5 |
+
- translation French English model
|
6 |
+
datasets:
|
7 |
+
- dcep europarl jrc-acquis
|
8 |
+
---
|
9 |
+
|
10 |
+
# legal_t5_small_trans_fr_en model
|
11 |
+
|
12 |
+
Pretrained model on protein sequences using a masked language modeling (MLM) objective. It was first released in
|
13 |
+
[this repository](https://github.com/agemagician/LegalTrans). This model is trained on three parallel corpus from jrc-acquis, europarl and dcep.
|
14 |
+
|
15 |
+
|
16 |
+
## Model description
|
17 |
+
|
18 |
+
legal_t5_small_trans_fr_en is based on the `t5-small` model and was trained on a large corpus of parallel text. This is a smaller model, which scales the baseline model of t5 down by using `dmodel = 512`, `dff = 2,048`, 8-headed attention, and only 6 layers each in the encoder and decoder. This variant has about 60 million parameters.
|
19 |
+
|
20 |
+
## Intended uses & limitations
|
21 |
+
|
22 |
+
The model could be used for translation of legal texts from French to English.
|
23 |
+
|
24 |
+
### How to use
|
25 |
+
|
26 |
+
Here is how to use this model to translate legal text from French to English in PyTorch:
|
27 |
+
|
28 |
+
```python
|
29 |
+
from transformers import AutoTokenizer, AutoModelWithLMHead, TranslationPipeline
|
30 |
+
|
31 |
+
pipeline = TranslationPipeline(
|
32 |
+
model=AutoModelWithLMHead.from_pretrained("SEBIS/legal_t5_small_trans_fr_en"),
|
33 |
+
tokenizer=AutoTokenizer.from_pretrained(pretrained_model_name_or_path = "SEBIS/legal_t5_small_trans_fr_en", do_lower_case=False,
|
34 |
+
skip_special_tokens=True),
|
35 |
+
device=0
|
36 |
+
)
|
37 |
+
|
38 |
+
fr_text = "(8) System vendors should ensure that CRS marketing data is available to all participating carriers without discrimination, and transport providers should not be able to use such data in order to unduly influence the choice of the travel agent nor the choice of the consumer .
|
39 |
+
"
|
40 |
+
|
41 |
+
pipeline([fr_text], max_length=512)
|
42 |
+
```
|
43 |
+
|
44 |
+
## Training data
|
45 |
+
|
46 |
+
The legal_t5_small_trans_fr_en model was trained on [JRC-ACQUIS](https://wt-public.emm4u.eu/Acquis/index_2.2.html), [EUROPARL](https://www.statmt.org/europarl/), and [DCEP](https://ec.europa.eu/jrc/en/language-technologies/dcep) dataset consisting of 5 Million parallel texts.
|
47 |
+
|
48 |
+
## Training procedure
|
49 |
+
|
50 |
+
### Preprocessing
|
51 |
+
|
52 |
+
### Pretraining
|
53 |
+
An unigram model with 88M parameters is trained over the complete parallel corpus to get the vocabulary (with byte pair encoding), which is used with this model.
|
54 |
+
|
55 |
+
|
56 |
+
## Evaluation results
|
57 |
+
|
58 |
+
When the model is used for translation test dataset, achieves the following results:
|
59 |
+
|
60 |
+
Test results :
|
61 |
+
|
62 |
+
| Model | secondary structure (3-states) |
|
63 |
+
|:-----:|:-----:|
|
64 |
+
| legal_t5_small_trans_fr_en | 51.44|
|
65 |
+
|
66 |
+
|
67 |
+
### BibTeX entry and citation info
|