Exploitation of Deep Learning and Neural Networks Techniques in Islamic Arabic Texts Processing

  • 28 May 2022
  • Recently published Research - Informatics & Communication


Mohammad Said Desouki

Published in

9th International Conference on Islamic Applications in Computer Science and Technologies (IMAN 2021).



There is a huge volume of Islamic Arabic texts in different sites and libraries around the world, and in different electronic forms. Some of them are in the form of images and PDF files, which need big efforts and long time to edit and to take advantage of their valuable information. In this format, information cannot be indexed in search engines to be retrieve it easily. Other texts exist in editable format but have very large volume so they need to be summarized or clustered to topics by their contents to be useful.

Today, many Deep Learning technics are used to train Neural Networks on data sets containing solved problems, to find later solutions to new problems provided by other data sets. Recently, many applications used these techniques in Natural Language Processing, and more specifically in Arabic Language Processing.

We try in this speech to make insights on how to take advantages of these new technics in Islamic Arabic Applications processing huge volumes of Islamic Arabic texts in different formats. We worked before on many researches in this domain using these technics, like researches on Arabic Text Summarization and Arabic Texts Authorship Attribution on which we published a paper in the previous IMAN conference titled “Simulating the Application of Authorship Attribution for Hadith Sharif Texts”.

We work now on building an Offline Handwritten Recognition system to recognize texts written in Historical Arabic Islamic Documents to transform it to transcripts of editable texts that can be indexed and searched easily, using an incremental learning approach and using a deep learning neural network that takes some pages that were recognized manually to learn how to recognize other pages.

Keywords: Deep Learning, Neural Networks, Arabic Text Processing, Offline Handwritten Recognition, Historical Arabic Islamic Documents.

Link to abstract