Personalized Response Generation via Generative Split Memory Network

Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 1956–1970 June 6–11, 2021. ©2021 Association for Computational Linguistics 1956 Personalized Response Generation via Generative Split Memory Network Yuwei Wu♣∗, Xuezhe Ma♠, Diyi Yang♦ ♣ Shanghai Jiao Tong University, will8821@sjtu.edu.cn ♠ University of Southern California, xuezhema@usc.edu ♦Georgia Institute of Technology, dyang888@gatech.edu Abstract Despite the impressive successes of generation and dialogue systems, how to endow a text generation system with particular personality traits to deliver more personalized responses remains under-investigated. In this work, we look at how to generate personalized responses for questions on Reddit by utilizing personal- ized user profiles and posting histories. Specif- ically, we release an open-domain single-turn dialog dataset made up of 1.5M conversation pairs together with 300k profiles of users and related comments. We then propose a memory network to generate personalized responses in dialogue that utilizes a novel mechanism of splitting memories: one for user profile meta attributes and the other for user-generated in- formation like comment histories. Experimen- tal results show the quantitative and qualitative improvements of our simple split memory net- work model over the state-of-the-art response generation baselines. The dataset and code are available here. 1 Introduction Building human-like conversational systems, in particular chit-chat agents, has been a long- standing goal in language technology communities. Unlike task-oriented dialog agents that focus on completing specific tasks (Wen et al., 2017; Eric et al., 2017; Lei et al., 2018; Lowe et al., 2015), chit-chat agents need to dynamically interact with people, understand the meaning of human conver- sations (Hovy and Yang, 2021), and thereby make better responses to improve user experience. Despite the recent successes on building chit- chat agents using data-driven approaches (Ritter et al., 2011; Banchs and Li, 2012; Serban et al., 2016; Li et al., 2016c; Parthasarathi and Pineau, 2018), lack of a consistent personality is still one of the common issues. The main reason is that these ∗The work is mainly done when YW was a visiting student at Georgia Institute of Technology. Question: Where do you live and what is something you are doing today? Responses: A: I live in Mongolia and I will be making some good sandwiches today. B: Midwest America, I will be skyping my brothers and going to band practice today. Question: What’s your "go to" when you’re sad? Responses: A: I listen to horror stories for some reason. B: I love to read or listen to sad music. Respondent Profile: A: Gender: female; Favorites: sandwich; Possessions: Russian class; Residence: Mongolia; Asia; B: Family: brothers; Self-description: guitarist; Favorite: fakebooks; Residence: America; Respondent Comment Histories: A: I often fall asleep while listening to horror stories. B: Listening to sad music, I know it adds fuel to fire but the flame will burn out quicker and you’ll feel better soon. Table 1: Example conversation pairs with respondents’ profile and posting histories, with related information from profile and histories in blue and red respectively. models are often trained …

Content truncated to 500 words