Towards Automatic Normalization of the Moroccan Dialectal Arabic User Generated Text

Today social media is an important way of communication between people in the world. As the case of other countries, Moroccan people use several languages in their web communication leaving behind a considerable amount of user-generated text. The latter presents several opportunities for extracting useful information. However, processing this content is very challenging especially when facing the Moroccan Dialectal Arabic content in social media. This is to several factors such as scripts diversity (Arabic and Arabizi), orthographic errors and writing rules lack. In this context, the present work is a first attempt towards addressing the problem of Moroccan Dialectal Arabic spelling inconsistency in social media. We conduct a deep study that uses a systematic approach where we report on a series of experiments performed on Moroccan Dialectal social media text. The most interesting findings that have emerged is the orthographic inconsistency existing in written Moroccan Dialectal Arabic regarding both Arabic and Latin scripts. This phenomenon affects an important amount of texts in social media and proved the need of exploiting available Arabic tools in addition to building a customized spelling correction system.