感情を模倣し 戦略を予測し応答⽣成 深い⽂脈理解や 認知的な応答⽣成が不⾜ 強化学習と戦略選択[4] 強化学習を⽤い ⽀援を求める⼈の反応に合わせ た最適な戦略を選択 認知的⽂脈理解が不⼗分 個別対応が困難 ペルソナを考慮した 戦略的デコーディング[5] ペルソナ情報を⽤いて 戦略的な応答を⽣成 認知⽂脈理解が不⾜し、複雑な 状況対応に限界 3 [2] Cheng, Y., Liu, W., Li, W., Wang, J., Zhao, R., Liu, B., ... & Zheng, Y. (2022). Improving multi-turn emotional support dialogue generation with lookahead strategy planning. arXiv preprint arXiv:2210.04242. [3] Sabour, S., Zheng, C., & Huang, M. (2022, June). Cem: Commonsense-aware empathetic response generation. In Proceedings of the AAAI Conference on Artificial Intelligence (Vol. 36, No. 10, pp. 11229-11237). [4] Cheng, Y., Liu, W., Li, W., Wang, J., Zhao, R., Liu, B., ... & Zheng, Y. (2022). Improving multi-turn emotional support dialogue generation with lookahead strategy planning. arXiv preprint arXiv:2210.04242.. [5] Rogers, A., Boyd-Graber, J., & Okazaki, N. (2023, July). Findings of the Association for Computational Linguistics: ACL 2023. In Findings of the Association for Computational Linguistics: ACL 2023.