Appendix of Stylized Dialogue Generation with Multi-Pass Dual Learning Jinpeng Li