Is there a way to determine a string is English or Arabic?
You could use N-gram-based text categorization (google for that phrase) but it is not a fail-proof technique, and it may require a not too short string.
You might also decide that a string with only ASCII letters is not Arabic.