An Analysis of Deep Reinforcement Learning Agents for Text-based Games

Text-based games(TBG) are complex environments which allow users or computer agents to make textual interactions and achieve game goals. It is challenging to build goal-oriented computer agents for text-based games, especially when we use step-wise feedback as the only text input for the model. Moreover, it is hard for agents to provide replies with flexible length and form by valuing from a much larger text input space. In this paper, we provide an extensive analysis of deep learning methods applied to the Text-Based Games field.