This document is about natural language processing (NLP), which focuses on developing effective communication systems between computers and humans. While the most significant advances in this area have been achieved through large language models (LLMs), these models often lack precision in rule-governed domains, such as spatial relations or legal norms. To address these domains, semantic parsers are used to assign logical representations to texts by analyzing their syntactic structure and semantic interpretation. However, these parsers are complex, and their design is challenging due to the manual implementation of specific rules. This study proposes an innovative approach using deep reinforcement learning to develop a semantic parser that can learn and adapt automatically. Through rewards, the agent will optimize its behavior over time, which could have a significant impact on the advancement of natural language processing.