QuestionAnswering : Qur'an Question Answering with Transformers

Transformers based approach for question answering in Qur'an which employs transfer-learning, ensemble-learning across multiple models.

Installation

You first need to install Java for the evaluation script which uses farasapy and the desired version is Java8. Please refer Oracle installation guide for more details on installing JDK for different platforms.

Then you need to install PyTorch. The recommended PyTorch version is 1.11.0 Please refer to PyTorch installation page for more details specifically for the platforms.

When PyTorch has been installed, you can install requirements from source by cloning the repository and running:

git clone https://github.com/DamithDR/QuestionAnswering.git
cd QuestionAnswering
pip install -r requirements.txt

Experiment Results

You can easily run experiments using following command and altering the parameters as you wish

python -m examples.arabic.quran.quran_question_answering --n_fold=1 --transfer_learning=False --self_ensemble=False --models=camelmix,arabert

Experiment using Docker

To run using docker, you need to have docker installed in your machine. Please use Docker installation Guide to install docker based on your operating system.

Once you successfully installed docker in your system, you can simply use following command to execute the experiments.

docker run damithpremasiri/question-answering-quran:v-1.0 --n_fold=1 --transfer_learning=False --self_ensemble=False --models=camelmix,arabert

Parameters

Please find the detailed descriptions of the parameters

n_fold              : Number of executions expected before self ensemble
transfer_learning   : On/Off transfer learning
self_ensemble       : On/Off self ensembling
models              : comma seperated model tags

Model Tags

arabert             : aubmindlab/bert-base-arabertv2
mbertcased          : bert-base-multilingual-cased
mbertuncased        : bert-base-multilingual-uncased
camelmix            : CAMeL-Lab/bert-base-arabic-camelbert-mix
camelca             : CAMeL-Lab/bert-base-arabic-camelbert-ca
araelectradisc      : aubmindlab/araelectra-base-discriminator
araelectragen       : aubmindlab/araelectra-base-generator

Citation

Please consider citing us if you use the library or the code.

@inproceedings{damith2022DTWquranqa,
  title={DTW at Qur'an QA 2022: Utilising Transfer Learning with Transformers for Question Answering in a Low-resource Domain},
  author={Damith Premasiri and Tharindu Ranasinghe and Wajdi Zaghouani and Ruslan Mitkov},
  booktitle={Proceedings of the 5th Workshop on Open-Source Arabic Corpora and Processing Tools (OSACT5).},
  year={2022}
}