APITO GPT: RAG-BASED ARTIFICIAL INTELLIGENCE SYSTEM FOR CONSULTATION AND INTERPRETATION OF OFFICIAL FOOTBALL RULES

In-depth knowledge of the rules of football is an indispensable condition for referees, assistant coaches and sports journalists, but it remains restricted to a single extensive official document that is difficult to consult contextually. This work presents the APITO GPT, a system of queries and answers based on the Retrieval-Augmented Generation (RAG) architecture, developed through the CRISP-DM methodology, which ingests, indexes and retrieves excerpts from the rulebook of the International Football Association Board (IFAB), adopted by the Brazilian Football Confederation (CBF), responding to queries in natural language with explicit normative grounding, confidence level and, when relevant, indexed images and tables of the work. The system shows high performance and rigor in factual correction, validating the technical feasibility and practical relevance for refereeing education and the popularization of football knowledge in Brazil.