The purpose of this study is to try to investigate the problem of deviation in non-standard data generated by inquiry into the Arabic language, using Artificial Intelligence (AI) or Natural Language Processing programs (NLP) applied to Arabic. Expert Systems and some Neural Networking used by AI could correct some deviation problems. We will try to find a model of analysis that could resolve some deviation cases in the Arabic language then generalizing it to other languages. This model is based on our linguistic and computational understanding of the phenomenon. The results should serve as a starting point for further research on deviant cases in general and the Arabic language in particular.