legal_t5_small_multitask_it_cs model

Model on translating legal text from Italian to Cszech. It was first released in this repository. The model is parallely trained on the three parallel corpus with 42 language pair from jrc-acquis, europarl and dcep along with the unsupervised task where the model followed the task of prediction in a masked language model.

Model description

No pretraining is involved in case of legal_t5_small_multitask_it_cs model, rather the unsupervised task is added with all the translation task to realize the multitask learning scenario.

Intended uses & limitations

The model could be used for translation of legal texts from Italian to Cszech.

How to use

Here is how to use this model to translate legal text from Italian to Cszech in PyTorch:

from transformers import AutoTokenizer, AutoModelWithLMHead, TranslationPipeline

pipeline = TranslationPipeline(
model=AutoModelWithLMHead.from_pretrained("SEBIS/legal_t5_small_multitask_it_cs"),
tokenizer=AutoTokenizer.from_pretrained(pretrained_model_name_or_path = "SEBIS/legal_t5_small_multitask_it_cs", do_lower_case=False, 
                                            skip_special_tokens=True),
    device=0
)

it_text = "Per mobilitare il Fondo, la Commissione ha presentato all'autorità di bilancio una richiesta di storno per un importo complessivo di 667.823 EUR dalla riserva FEG (40 02 43) in stanziamenti d'impegno verso la linea di bilancio FEG."

pipeline([it_text], max_length=512)

Training data

The legal_t5_small_multitask_it_cs model (the supervised task which involved only the corresponding langauge pair and as well as unsupervised task where all of the data of all language pairs were available) model was trained on JRC-ACQUIS, EUROPARL, and DCEP dataset consisting of 5 Million parallel texts.

Training procedure

The model was trained on a single TPU Pod V3-8 for 250K steps in total, using sequence length 512 (batch size 4096). It has a total of approximately 220M parameters and was trained using the encoder-decoder architecture. The optimizer used is AdaFactor with inverse square root learning rate schedule.

Preprocessing

An unigram model trained with 88M lines of text from the parallel corpus (of all possible language pairs) to get the vocabulary (with byte pair encoding), which is used with this model.

Pretraining

Evaluation results

When the model is used for translation test dataset, achieves the following results:

Test results :

Model	BLEU score
legal_t5_small_multitask_it_cs	37.935

BibTeX entry and citation info

Created by Ahmed Elnaggar/@Elnaggar_AI | LinkedIn