|
--- |
|
language: ar |
|
datasets: |
|
- wikipedia |
|
- Osian |
|
- 1.5B-Arabic-Corpus |
|
- oscar-arabic-unshuffled |
|
- Assafir(private) |
|
--- |
|
|
|
# AraELECTRA |
|
|
|
<img src="https://raw.githubusercontent.com/aub-mind/arabert/master/AraELECTRA.png" width="100" align="left"/> |
|
|
|
**ELECTRA** is a method for self-supervised language representation learning. It can be used to pre-train transformer networks using relatively little compute. ELECTRA models are trained to distinguish "real" input tokens vs "fake" input tokens generated by another neural network, similar to the discriminator of a [GAN](https://arxiv.org/pdf/1406.2661.pdf). AraELECTRA achieves state-of-the-art results on Arabic QA dataset. |
|
|
|
For a detailed description, please refer to the AraELECTRA paper [AraELECTRA: Pre-Training Text Discriminators for Arabic Language Understanding](https://arxiv.org/abs/2012.15516). |
|
|
|
## How to use the discriminator in `transformers` |
|
|
|
```python |
|
from transformers import ElectraForPreTraining, ElectraTokenizerFast |
|
import torch |
|
|
|
discriminator = ElectraForPreTraining.from_pretrained("aubmindlab/araelectra-base-discriminator") |
|
tokenizer = ElectraTokenizerFast.from_pretrained("aubmindlab/araelectra-base-discriminator") |
|
|
|
sentence = "" |
|
fake_sentence = "" |
|
|
|
fake_tokens = tokenizer.tokenize(fake_sentence) |
|
fake_inputs = tokenizer.encode(fake_sentence, return_tensors="pt") |
|
discriminator_outputs = discriminator(fake_inputs) |
|
predictions = torch.round((torch.sign(discriminator_outputs[0]) + 1) / 2) |
|
|
|
[print("%7s" % token, end="") for token in fake_tokens] |
|
|
|
[print("%7s" % int(prediction), end="") for prediction in predictions.tolist()] |
|
``` |
|
|
|
# Model |
|
|
|
Model | HuggingFace Model Name | Size (MB/Params)| |
|
---|:---:|:---: |
|
AraELECTRA-base-generator | [araelectra-base-generator](https://huggingface.co/aubmindlab/araelectra-base-generator) | 227MB/60M | |
|
AraELECTRA-base-discriminator | [araelectra-base-discriminator](https://huggingface.co/aubmindlab/araelectra-base-discriminator) | 516MB/135M | |
|
|
|
# Compute |
|
Model | Hardware | num of examples (seq len = 512) | Batch Size | Num of Steps | Time (in days) |
|
---|:---:|:---:|:---:|:---:|:---: |
|
AraELECTRA-base | TPUv3-8 | - | 256 | 2M | 24 |
|
|
|
# Dataset |
|
|
|
The pretraining data used for the new **AraELECTRA** model is also used for **AraGPT2 and AraBERTv2**. |
|
|
|
The dataset consists of 77GB or 200,095,961 lines or 8,655,948,860 words or 82,232,988,358 chars (before applying Farasa Segmentation) |
|
|
|
For the new dataset we added the unshuffled OSCAR corpus, after we thoroughly filter it, to the previous dataset used in AraBERTv1 but with out the websites that we previously crawled: |
|
- OSCAR unshuffled and filtered. |
|
- [Arabic Wikipedia dump](https://archive.org/details/arwiki-20190201) from 2020/09/01 |
|
- [The 1.5B words Arabic Corpus](https://www.semanticscholar.org/paper/1.5-billion-words-Arabic-Corpus-El-Khair/f3eeef4afb81223df96575adadf808fe7fe440b4) |
|
- [The OSIAN Corpus](https://www.aclweb.org/anthology/W19-4619) |
|
- Assafir news articles. Huge thank you for Assafir for giving us the data |
|
|
|
# Preprocessing |
|
|
|
It is recommended to apply our preprocessing function before training/testing on any dataset. |
|
|
|
**Install the arabert python package to segment text for AraBERT v1 & v2 or to clean your data `pip install arabert`** |
|
|
|
```python |
|
from arabert.preprocess import ArabertPreprocessor |
|
|
|
model_name="araelectra-base" |
|
arabert_prep = ArabertPreprocessor(model_name=model_name) |
|
|
|
text = "ููู ูุจุงูุบ ุฅุฐุง ูููุง ุฅู ูุงุชู ุฃู ูู
ุจููุชุฑ ุงูู
ูุชุจ ูู ุฒู
ููุง ูุฐุง ุถุฑูุฑู" |
|
arabert_prep.preprocess(text) |
|
|
|
>>> output: ููู ูุจุงูุบ ุฅุฐุง ูููุง : ุฅู ูุงุชู ุฃู ูู
ุจููุชุฑ ุงูู
ูุชุจ ูู ุฒู
ููุง ูุฐุง ุถุฑูุฑู |
|
``` |
|
|
|
|
|
# TensorFlow 1.x models |
|
|
|
**You can find the PyTorch, TF2 and TF1 models in HuggingFace's Transformer Library under the ```aubmindlab``` username** |
|
|
|
- `wget https://huggingface.co/aubmindlab/MODEL_NAME/resolve/main/tf1_model.tar.gz` where `MODEL_NAME` is any model under the `aubmindlab` name |
|
|
|
# If you used this model please cite us as : |
|
|
|
``` |
|
@inproceedings{antoun-etal-2021-araelectra, |
|
title = "{A}ra{ELECTRA}: Pre-Training Text Discriminators for {A}rabic Language Understanding", |
|
author = "Antoun, Wissam and |
|
Baly, Fady and |
|
Hajj, Hazem", |
|
booktitle = "Proceedings of the Sixth Arabic Natural Language Processing Workshop", |
|
month = apr, |
|
year = "2021", |
|
address = "Kyiv, Ukraine (Virtual)", |
|
publisher = "Association for Computational Linguistics", |
|
url = "https://www.aclweb.org/anthology/2021.wanlp-1.20", |
|
pages = "191--195", |
|
} |
|
``` |
|
|
|
# Acknowledgments |
|
Thanks to TensorFlow Research Cloud (TFRC) for the free access to Cloud TPUs, couldn't have done it without this program, and to the [AUB MIND Lab](https://sites.aub.edu.lb/mindlab/) Members for the continous support. Also thanks to [Yakshof](https://www.yakshof.com/#/) and Assafir for data and storage access. Another thanks for Habib Rahal (https://www.behance.net/rahalhabib), for putting a face to AraBERT. |
|
|
|
# Contacts |
|
**Wissam Antoun**: [Linkedin](https://www.linkedin.com/in/wissam-antoun-622142b4/) | [Twitter](https://twitter.com/wissam_antoun) | [Github](https://github.com/WissamAntoun) | <wfa07@mail.aub.edu> | <wissam.antoun@gmail.com> |
|
|
|
**Fady Baly**: [Linkedin](https://www.linkedin.com/in/fadybaly/) | [Twitter](https://twitter.com/fadybaly) | [Github](https://github.com/fadybaly) | <fgb06@mail.aub.edu> | <baly.fady@gmail.com> |
