The problem of entity factorizing versus unfactorizing is one of the main
problems that face peoples in the human languages technologies field. As a case
study for this problem; this thesis studies the automatic Arabic text diacritization
problem. The thesis compares the diacritization through words factorization
using morphological analyses versus the diacritization through the words
unfactorization using full-form words. From the experimental results show that;
for small training corpus size; unfactorizing system is better since it can reach the
saturation faster than the factorizing one, but it may suffer from the OOV
problem, but for very large training corpus size; the two systems are almost the
same, except that the cost of the unfactorizing systems is lower. So, the best
strategy is using a hybrid of the two systems to enjoy the fast learning and low
cost of the factorizing system and the wide coverage of the factorizing one.