Text this: A Reward Optimization Method Based on Action Subrewards in Hierarchical Reinforcement Learning