Detalhes do Documento

This paper presents an evaluation of large language models (LLMs) applied to the task of normalizing eighteenth-century written texts. Several LLMs were employed to process texts in pre-contemporary spellings and update them according to contemporary Portuguese orthography. Their outputs were rigorously compared against a curated reference corpus. The findings indicate marked disparities in model performance, with the Portuguese specialized model Sabiá demonstrating a statistically significant advantage over multilingual alternatives.

This work is funded by national funds through the Foundation for Science and Technology (FCT - Portugal), under the project UID/00057/2025, partially financed by CNPq (National Council for Scientific and Technological Development), grant 307088/2023-5, FAPERJ - Fundação Carlos Chagas Filho de Amparo à Pesquisa do Estado do Rio de Janeiro, processes SEI-260003/002930/2024, SEI-260003/000614/2023, and Coordenação de Aperfeiçoamento de Pessoal de Nível Superior – Brasil (CAPES) – Finance Code 001. We also thank the support of the CNPq National Institutes of Science and Technology, IAIA (grant 406417/2022-9), TILD-IAR (grant 408490/2024-1).