Adel Atef, Bassam Mattar, Sandra Sherif, Eman Elrefai, Marwan Torki
Alexandria University, Egypt
[email protected], [email protected], [email protected], [email protected] ,[email protected]
Current Arabic Machine Reading for Question Answering datasets suffers from important shortcomings. The available datasets are either small-sized high-quality collection or large-sized low-quality dataset. To address the aforementioned problem we present our Arabic Question-Answer dataset (AQAD). AQAD is a new Arabic reading comprehension large-sized high-quality dataset consisting of 17,000+ questions. To collect the AQAD dataset, we present a fully automated data collector. Our collector works on a set of Arabic Wikipedia articles for extractive question answering task. The chosen articles match the articles used in the well-known SQuAD dataset. We provide evaluation results on the AQAD dataset using two state-of-the-art models for machine-reading question answering problems Namely, BERT and BIDAF models which result in 0.37 and 0.32 F-1 measure on AQAD dataset.
Automated dataset collection, Arabic Machine Comprehension, Question Answering System, Arabic Machine Reading for Question Answering (A-MRQA), Arabic NLP, Answer Extraction.
The project is divided into 2 main phases:
We implemented an automatic dataset generator system capapble of generating arabic dataset with question-answer pairs.
We plugged our dataset to known baseline models to evaluate some interesting results.
To reprocuce results given by the paper, you just need to open any required .ipynb in colab and run the cell in order from top to bottom to install requirements.