Arabic script character recognition is challenging task due to complexity of the script and huge number of ligatures. We present a method for the development of multilingual Arabic script OCR (Optical Character Recognition) and lexicon reduction for Arabic Script and its derivative languages. The objective of the proposed method is to overcome the large dataset Urdu and similar scripts by using GCT (Ghost Character Theory) concept. Arabic and its sibling script languages share the similar character dataset i.e. the character set are difference in diacritic and writing styles like Naskh or Nasta?liq. Based on the proposed method, the lexicon for Arabic and Arabic script based languages can be minimized approximately up to 20 times. The proposed multilingual Arabic script OCR approach have been evaluated for online Arabic and its derivative language like Urdu using BPNN. The result showed that proposed method helps to not only the reduction of lexicon but also helps to develop the Multilanguage character recognition system for Arabic Script