Aprimorando o atendimento ao cidadão com RAGLC (Retrieval Augmented Generation in Logic Context), uma nova abordagem RAG

Abstract

This work develops and evaluates an intelligent conversational assistant, based on Retrieval Augmented Generation (RAG) and Large Language Models (LLMs), to automate and improve service to candidates for public examinations and selection processes, as well as new hires at the Uberlândia City Hall. It addresses concepts based on Natural Language Processing (NLP) and low-cost Artificial Intelligence for the Agency. The growing demand and supply of services by the City Hall, and the increasing number of turnover among employees, means that the number of municipal hires and the number of civil servants is increasing. This movement directly overloads the civil servants in the personnel recruitment areas, which, despite great effort, generates dissatisfaction in the service provided by many candidates and newcomers to the municipal staff. This entire context motivated the creation of a solution that increases the efficiency, precision and speed of this public service. The methodology employs an advanced RAG architecture, which stands out for proposing an innovative text segmentation technique based on semantic coherence, outperforming conventional fixed-length division methods. A rigorous empirical analysis was performed, comparing multiple retrieval strategies, including the enrichment of user questions promoted by LLM, and evaluating the impact of different embedding models on the quality of the answers. The expected results include the evaluation of an efficient model capable of processing demands from more than 80,000 candidates who took part in the public examinations and selection processes of the City Hall of Uberlândia in the last 4 years, demonstrating how automation combined with Artificial Intelligence can improve the quality of public service. It is concluded that the strategic use of AI technologies, specifically RAG systems with optimized context recovery, can not only improve administrative processes, but also establish a new paradigm of excellence, transparency, and accessibility in public service.