Shakkala Project V 1.0 مشروع شكّالة

Introduction

Shakkala project use recurrent neural network to automatically form Arabic characters (تشكيل الحروف).
This is beta version 1.0 with big improve from previous version 0.1

Requirements

Execute following commands:

cd requirements
pip install -r requirements.txt

Code Examples (How to)

Check full example in (demo.py) file.

Create Shakkala object

sh = Shakkala(folder_location)

Prepare input

input_int = sh.prepare_input(input_text)

Call the neural network

model, graph = sh.get_model()
with graph.as_default():
      logits = model.predict(input_int)[0]

Predict output

predicted_harakat = sh.logits_to_text(logits)
final_output = sh.get_final_text(input_text, predicted_harakat)

Accuracy

In this version accuracy reached almost 90% and in some data it reach more. This beta version trained on personal computer with majority of historical Arabic data from books and some of available formed modern data in the internet.

Prediction Example

Following results predicted with model (middle_model.h5)

Example1

Real output :
فَإِنْ لَمْ يَكُونَا كَذَلِكَ أَتَى بِمَا يَقْتَضِيهِ الْحَالُ وَهَذَا أَوْلَى
Predicted output :
فَإِنْ لَمْ يَكُونَا كَذَلِكَ أَتَى بِمَا يَقْتَضِيهِ الْحَالَ وَهَذَا أَوْلَى

Example2

Real output :
قَالَ الْإِسْنَوِيُّ وَسَوَاءٌ فِيمَا قَالُوهُ مَاتَ فِي حَيَاةِ أَبَوَيْهِ أَمْ لَا
Predicted output :
قَالَ الْإِسْنَوِيُّ وَسَوَاءٌ فِيمَا قَالُوهُ مَاتَ فِي حَيَاةِ أَبَوَيْهِ أَمْ لَا

Accuracy Enhancements

The model can be enhanced to reach 95%-98% accuracy with following:

Availability of modern formed data to train the network. (because current version trained with mostly available historical Arabic data and some modern data)
Rent a server with high GPU to increase the number of epochs, layer and neural units.

Model Design

Team

Ahmad Barqawi: Neural Network Developer.
Taha Zerrouki: Mentor Data and Results.

License

Free to use and distribute only mention the original project name Shakkala as base model. The MIT License (MIT)

Copyright (c) 2017 Shakkala Project

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

ihabalnaqib / Shakkala