Reinforcement Learning for Personalized Dialogue Management

TL;DR

This paper addresses personalized dialogue management (DM) by introducing two reinforcement learning (RL) approaches that generalize user context across multiple users. It formalizes the DM task within a POMDP framework and investigates segmentation-based versus state-based context integration, evaluating them on an extended benchmark that adds a financial product recommendation domain. The results show that learning-based DM often outperforms handcrafted baselines, especially when user context and domain variations are present, though effectiveness depends on domain, data availability, and algorithm choice; context can improve personalization even without prior interactions. Overall, the work demonstrates feasibility and value of context-aware RL for DM and provides a benchmark and methodological groundwork for future studies in personalized conversational systems, including more complex contexts and broader domains.

Abstract

Language systems have been of great interest to the research community and have recently reached the mass market through various assistant platforms on the web. Reinforcement Learning methods that optimize dialogue policies have seen successes in past years and have recently been extended into methods that personalize the dialogue, e.g. take the personal context of users into account. These works, however, are limited to personalization to a single user with whom they require multiple interactions and do not generalize the usage of context across users. This work introduces a problem where a generalized usage of context is relevant and proposes two Reinforcement Learning (RL)-based approaches to this problem. The first approach uses a single learner and extends the traditional POMDP formulation of dialogue state with features that describe the user context. The second approach segments users by context and then employs a learner per context. We compare these approaches in a benchmark of existing non-RL and RL-based methods in three established and one novel application domain of financial product recommendation. We compare the influence of context and training experiences on performance and find that learning approaches generally outperform a handcrafted gold standard.

Figures (3)

• Figure 1: RL-based approaches to personalized DM.
• Figure 2: Average reward per dialogue in test set for environments without (a) and with (b) ASR/NLU errors.
• Figure 3: Per-dialogue reward of selected algorithms in test set, averaged over all environments.