Uso de modelos de linguagem para catalogação automatizada de cartas de sesmaria: uma aplicação à plataforma SILB

Abstract

This study investigates the application of Large Language Models (LLMs) to automate the cataloging of colonial land grant documents (sesmarias) from the SILB Platform, addressing challenges posed by archaic language, textual variations, and lack of standardization. Using GPT-4o and advanced prompt engineering techniques (zero-shot, few-shot, and chain-of-thought), the system extracts structured data (grantee names, dates, geographic locations) and identifies inconsistencies in existing records. Results showed effectiveness in objective fields (60% accuracy) but limitations in contextual analysis (e.g., justifications, observations). The research advances digital preservation of historical archives by reducing cataloging time and costs, while critically examining AI’s boundaries in processing non-standardized primary sources.