Generate, Delete and Rewrite: A Three-Stage Framework for Improving Persona Consistency of Dialogue Generation

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 5821–5831 July 5 - 10, 2020. c⃝2020 Association for Computational Linguistics 5821 Generate, Delete and Rewrite: A Three-Stage Framework for Improving Persona Consistency of Dialogue Generation Haoyu Song1, Yan Wang2, Wei-Nan Zhang1∗, Xiaojiang Liu2, Ting Liu1 1Research Center for Social Computing and Information Retrieval Harbin Institute of Technology, Heilongjiang, China 2 Tencent AI Lab, Shenzhen, China {hysong,wnzhang,tliu}@ir.hit.edu.cn {brandenwang,kieranliu}@tencent.com Abstract Maintaining a consistent personality in con- versations is quite natural for human beings, but is still a non-trivial task for machines. The persona-based dialogue generation task is thus introduced to tackle the personality- inconsistent problem by incorporating explicit persona text into dialogue generation mod- els. Despite the success of existing persona- based models on generating human-like re- sponses, their one-stage decoding framework can hardly avoid the generation of incon- sistent persona words. In this work, we introduce a three-stage framework that em- ploys a generate-delete-rewrite mechanism to delete inconsistent words from a generated re- sponse prototype and further rewrite it to a personality-consistent one. We carry out eval- uations by both human and automatic metrics. Experiments on the Persona-Chat dataset show that our approach achieves good performance. 1 Introduction In an open-domain conversation scenario, two speakers conduct open-ended chit-chat from the initial greetings and usually come to focus on their characteristics, such as hobbies, pets, and occupa- tions, etc., in the course of the conversation. For humans, they can easily carry out conversations according to their personalities (Song et al., 2019a), but fulfilling this task is still a challenge for recent neural dialogue models (Welleck et al., 2019). One main issue is that these models are typically trained over millions of dialogues from different speakers, and the neural dialogue models have a propensity to mimic the response with the maxi- mum likelihood in the training corpus (Li et al., 2016b), which results in the frequent inconsistency in responses (Zhang et al., 2018). Another issue ∗This work was done when the first author was an intern at Tencent AI Lab. Wei-Nan Zhang is the corresponding author. I’m a recording engineer; I live in California Hi, Kevin here. I love Mexican food. Hi I am Tom. I am in Colorado. Where do you live? Hi I am Tom. <mask> <mask> <mask><mask>. Where do you live? Hi I am Tom. I’m an engineer in California. Where do you live? Generate Delete Rewrite Inconsistent Personality Stage 1: Stage 2: Stage 3: Query: Persona: Figure 1: A common problem for persona-based dia- logue models is that they can hardly avoid the genera- tion of inconsistent persona words. Although the model generates a response which looks good, it is an incon- sistent one. With further rewriting, the model can focus more on improving persona consistency. is the user-sparsity problem (Qian et al., 2017) in conventional dialogue corpora (Serban et al., 2015). Some users have very few dialogue data, which makes it difficult for neural models to learn mean- ingful user representations (Li et al., …

Content truncated to 500 words