Reinforcement learning (RL) has emerged as a powerful framework for solving sequential decision-making problems in various domains, including speech and language technologies. This chapter delves into the formulation of RL problems in the context of speech and language applications. We explore how Markov decision processes (MDPs) can be used to model these applications and discuss the key components of RL formulations, such as the state space, action space, and reward functions. Additionally, we highlight the importance of metrics and evaluation strategies specific to RL in speech and language tasks. By understanding these formulations and evaluation methods, researchers and practitioners can effectively apply RL techniques to improve the performance and adaptability of speech and language systems.