Nunes, Isabel DillmannNacimento, Jefferson Rodrigues do2024-04-262024-04-262024-02-28NASCIMENTO, Jefferson Rodrigues. Exploração de técnicas de engenharia de prompt para aprimorar os resultados do uso de LLM no TCMRio. 2024. 60 f. Trabalho de Conclusão de Curso (Especialização em Residência em Tecnologia da Informação) - Instituto Metrópole Digital, Universidade Federal do Rio Grande do Norte, Natal, 2024.https://repositorio.ufrn.br/handle/123456789/58251The adoption of Large Language Models (LLMs) by the Court of Auditors of the Municipality of Rio de Janeiro (TCMRio) signifies a significant leap in its technological innovation journey, aiming to enhance internal operations through Artificial Intelligence. The motivation for this study stems from a proof of concept conducted by TCMRio, which uncovered substantial limitations in the quality of automated interactions, highlighting the need to explore new approaches to optimize these outcomes. In this context, the research seeks to investigate prompt engineering techniques as a solution to overcome the identified challenges, thereby enhancing the effectiveness of the employed LLMs. The specific objectives include a theoretical review of the fundamentals of Artificial Intelligence (AI), Natural Language Processing (NLP), and prompt engineering, with an emphasis on improving the use of LLMs, particularly GPT, within the TCMRio environment. Additionally, it proposes the practical exploration of these techniques through the development and testing of prompts aimed at improving the quality and relevance of chatbot responses. The adopted methodology features a mixed approach, combining detailed bibliographic analysis and practical application. This involves the implementation of a chatbot prototype using LLM, followed by a series of iterative tests to refine the prompts based on the studied prompt engineering techniques. The evaluation of results focused on the efficiency and effectiveness of the chatbot's responses, comparing them to those from the initial proof of concept. The findings demonstrate the viability and effectiveness of prompt engineering techniques in significantly improving the quality of interactions with the chatbot. The developed strategies contribute to a promising path for the future implementation of AI-based solutions that meet the specific needs of TCMRio.Attribution 3.0 Brazilhttp://creativecommons.org/licenses/by/3.0/br/Engenharia de PromptInteligência ArtificialProcessamento de Linguagem NaturalLangChainPrompt EngineeringArtificial IntelligenceNatural Language ProcessingExploração de técnicas de engenharia de prompt para aprimorar os resultados do uso de LLM no TCMRioExploring Prompt Engineering Techniques to Enhance LLM Results in TCMRiobachelorThesisCNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO