# access

**Repository Path**: facebookresearch/access

## Basic Information

- **Project Name**: access
- **Description**: Code to reproduce the experiments from the paper.
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: dependabot/pip/joblib-1.2.0
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2023-07-24
- **Last Updated**: 2024-10-23

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# Controllable Sentence Simplification

This repository contains the original implementation of the ACCESS model (**A**udien**C**e-**CE**ntric **S**entence **S**implification)  presented in [Controllable Sentence Simplification](https://arxiv.org/abs/1910.02677).

The version that was used at submission time is on branch [submission](https://github.com/facebookresearch/access/tree/submission).


## Getting Started

### Dependencies

* Python 3.6

### Installing

```
git clone git@github.com:facebookresearch/access.git
cd access
pip install -e .
pip install --force-reinstall easse@git+git://github.com/feralvam/easse.git@580ec953e4742c3ae806cc85d867c16e9f584505
pip install --force-reinstall fairseq@git+https://github.com/louismartin/fairseq.git@controllable-sentence-simplification

pip install torch==1.2
pip install sacrebleu==1.3.7
```

### How to use

Evaluate the pretrained model on WikiLarge:
```
python scripts/evaluate.py
```

Simplify text with the pretrained model
```
python scripts/generate.py < my_file.complex
```

Train a model
```
python scripts/train.py
```

## Pretrained model

The fairseq checkpoint of our model with the best scores can be found [here](http://dl.fbaipublicfiles.com/access/best_model.tar.gz).
The model's output simplifications can be viewed on the [EASSE HTML report](http://htmlpreview.github.io/?https://github.com/facebookresearch/access/blob/master/system_output/easse_report.html).

## References

If you use this code, please cite:  
L. Martin, B. Sagot, E. De la Clergerie, A. Bordes, [*Controllable Sentence Simplification*](https://arxiv.org/abs/1910.02677)

## Author

If you have any question, please contact the author:
**Louis Martin** ([louismartincs@gmail.com](mailto:louismartincs@gmail.com))

## License

See the [LICENSE](LICENSE) file for more details.