Text this: Leveraging Graph Networks to Model Environments in Reinforcement Learning