You have a typical pass-through configuration scenario: data in another character set stored under incorrect character set declaration (WE8PC850 does not formally encode any Arabic characters!). Export/import cannot help repairing such situation. You need to correct the declaration of the original database by changing the character set declaration (without touching data!) from the incorrect WE8PC850 to the actual character set of the data, which is one of the Arabic character sets. We need to find out first what this correct character set is. The easiest way to find out is to analyze the client applications that stored the data in the first place. Can you tell us something about those applications, especially the technology they are based on and the operating system platform on which they run?