A Comparative Study of Some Automatic Arabic Text Diacritization Systems

Table 2

Some statistics about the corpus splits used in this study.

	Train set	Validation set	Test set
Characters	16082164	784570	820022
Tokens	2460405	120075	125220
Numbers	33260	1648	1637
Digits	75963	3794	3774
Arabic words	2103156	102479	107291
Arabic letters	8356030	407434	426469
Diacritics	7290312	355666	371726
Undiacritized forms	105720	19515	20520
Diacritized forms	163237	26129	27298