Paper: Hierarchical Reinforcement Learning for Adaptive Text Generation

ACL ID W10-4204
Title Hierarchical Reinforcement Learning for Adaptive Text Generation
Venue International Conference on Natural Language Generation
Session Main Conference
Year 2010
Authors

We present a novel approach to natural lan- guage generation (NLG) that applies hierar- chical reinforcement learning to text genera- tion in the wayfinding domain. Our approach aims to optimise the integration of NLG tasks that are inherently different in nature, such as decisions of content selection, text struc- ture, user modelling, referring expression gen- eration (REG), and surface realisation. It also aims to capture existing interdependen- cies between these areas. We apply hierar- chical reinforcement learning to learn a gen- eration policy that captures these interdepen- dencies, and that can be transferred to other NLG tasks. Our experimental results—in a simulated environment—show that the learnt wayfinding policy outperforms a baseline pol- icy that takes reasonable actio...