A New Multi Varied Arabic Corpus

Khouloud Meskaldji, Salim Chikhi, Imene Bensalem

Abstract: This paper presents a new Arabic text corpus for Arabic variety detection which includes more than 233.000 words built from three varied sources of Arabic language: Quranic text, Classical Arabic text, and Modern Arabic text. Characteristics of the corpus and the steps of construction are presented and the corpus will be soon freely available for download and use after improvements.