Identifying CDRs in Large Data Sets