Automatic restoration of Arabic diacritics : a simple, purely statistical approach

Abstract EN

This paper is about an innovative Arabic discretize.

Most of the previous published works on discretization depends on other tools such as Hidden Markov Model Toolkit and / or higher linguistic knowledge such as morphology and syntax.

The present system is a baseline system which is small in size, fast in processing, and independent from other tools and linguistic rules.

The system uses a statistical method that relies on quad-gram probabilities.

Its accuracy rate is relatively high when compared to previous systems that are based only on statistics.