On inferring zero-reversible languages