Transferable Persona-Grounded Dialogues via Grounded Minimal Edits

Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 2368–2382 November 7–11, 2021. c⃝2021 Association for Computational Linguistics 2368 Transferable Persona-Grounded Dialogues via Grounded Minimal Edits Chen Henry Wu1, Yinhe Zheng2, Xiaoxi Mao3, Minlie Huang1 1 Department of Computer Science and Technology, Institute for Artificial Intelligence, State Key Lab of Intelligent Technology and Systems, Beijing National Research Center for Information Science and Technology, Tsinghua University, Beijing, China 2 Samsung Research China - Beijing (SRC-B) 3 Fuxi AI Lab, NetEase Inc., Hangzhou, China henrychenwu@cmu.edu, yh.zheng@samsung.com, maoxiaoxi@corp.netease.com, aihuang@tsinghua.edu.cn Abstract Grounded dialogue models generate responses that are grounded on certain concepts. Limited by the distribution of grounded dialogue data, models trained on such data face the transfer- ability challenges in terms of the data distribu- tion and the type of grounded concepts. To ad- dress the challenges, we propose the grounded minimal editing framework, which minimally edits existing responses to be grounded on the given concept. Focusing on personas, we pro- pose Grounded Minimal Editor (GME), which learns to edit by disentangling and recombin- ing persona-related and persona-agnostic parts of the response. To evaluate persona-grounded minimal editing, we present the PERSONAMI- NEDIT dataset, and experimental results show that GME outperforms competitive baselines by a large margin. To evaluate the transferabil- ity, we experiment on the test set of BLEND- EDSKILLTALK and show that GME can edit dialogue models’ responses to largely improve their persona consistency while preserving the use of knowledge and empathy.1 1 Introduction Grounding dialogue agents on external information is important for building engaging conversational AI systems (Huang et al., 2020). Along this track, various datasets and models have been proposed to ground dialogues on personas (Zhang et al., 2018), knowledge (Dinan et al., 2019), emotions (Zhou et al., 2018a), and images (Shuster et al., 2020). Generally, grounded dialogue modeling trains a dialogue model on a dataset D that consists of triples (c, r, g), where c is the dialogue history, r is the response, and g is the grounded concept. The model is generally optimized using maximum likelihood estimate (MLE), i.e., arg max θ E(c,r,g)∼D log Pθ(r|c, g). (1) 1Our codes and data are available at https://github. com/thu-coai/grounded-minimal-edit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…

Content truncated to 500 words