On the Bi-gram Model Based Auto Proofreading of Non-word Errors in Kazakh Texts