Automated Prediction of Medieval Arabic Diacritics

TL;DR

This paper presents a neural machine translation approach using LSTM-based bi-directional RNNs for diacritization of Medieval Arabic, improving upon existing tools and providing an accessible Python package.

Contribution

It introduces a novel neural model for Arabic diacritization and emphasizes the importance of context size in optimizing prediction accuracy.

Findings

Model outperforms baseline online tool

Published diacritization model as an open-source Python package

Highlighting the significance of context size in model optimization

Abstract

This study uses a character level neural machine translation approach trained on a long short-term memory-based bi-directional recurrent neural network architecture for diacritization of Medieval Arabic. The results improve from the online tool used as a baseline. A diacritization model have been published openly through an easy to use Python package available on PyPi and Zenodo. We have found that context size should be considered when optimizing a feasible prediction model.