An Overview of Natural Language State Representation for Reinforcement Learning
This is an incremental survey paper addressing the problem of state representation for researchers in reinforcement learning and natural language processing.
The paper surveys strategies for constructing natural language state representations in reinforcement learning, advocating for more interpretable and grounded designs with careful evaluation.
A suitable state representation is a fundamental part of the learning process in Reinforcement Learning. In various tasks, the state can either be described by natural language or be natural language itself. This survey outlines the strategies used in the literature to build natural language state representations. We appeal for more linguistically interpretable and grounded representations, careful justification of design decisions and evaluation of the effectiveness of different approaches.