Identifying main obstacles for statistical machine translation of morphologically rich South Slavic languages