Students and teachers create ‘Portinari’s Linguistic Signature’ project

A team of students and professors from IMPA Tech is using Artificial Intelligence (AI) tools to investigate a little-explored aspect of Candido Portinari’s (1903-1962) work: his writing. The project “ Portinari’s Linguistic Signature“, developed in partnership with the PUC-Behring Institute for Artificial Intelligence as part of the Portinari ProjectProject, aims to analyze the artist’s collection of personal letters to identify linguistic patterns capable of revealing traces of authorship and aspects of his thinking.

Recognized as one of Brazil’s greatest painters and a central figure of Modernism in the country, Portinari’s works depicted themes such as social inequality, life in the countryside, childhood and the suffering of workers. The letters written by the artist are a rich source for cultural and linguistic studies, allowing us to better understand his worldview, his creative processes and the historical and social context in which he lived.

Using this material, the project aims to identify linguistic artifacts, lexical choices and syntactic patterns that characterize the painter’s so-called linguistic signature. The aim is to develop methods capable of deciding whether certain passages were actually written by Portinari, thus contributing to the identification of textual authorship.

To do this, the team uses advanced AI techniques, especially Large Scale Language Models (LLMs) and foundational models. The development of new methodologies for organizing and analyzing linguistic data could, in the future, be applied to different corpora – data sets made up of digital or digitized linguistic resources, annotated or unannotated. This type of approach makes it possible to extract psychological and socio-cultural characteristics of individuals, populations and historical periods.

The project was created in partnership with João Candido Portinari, the painter’s son and general director of the Portinari Project. At IMPA Tech, professors Emilio Brazil, Cilene Rodrigues, Asla Sá, Rodrigo Ribeiro and Uéverton Souza are participating in the initiative, each responsible for a stage of the research and for guiding the students in the design, development and production of the material.

The Technology and Innovation Mathematics students are in the initial phase of their work, dedicated to curating the collection and analyzing the letters written by Portinari.

“It’s been a great experience. I’ve always loved art and literature. With the project, as well as learning more about Portinari, I’m able to combine mathematics and these passions. It’s exciting, because everyone knows about the famous painter who changed Brazilian art, but now we’re seeing who he was and what he wrote,” says student Nicole Freire.

On February 26, the students went to the Institute of Letters at PUC-Rio (Pontifical Catholic University of Rio de Janeiro) for a meeting with João Candido, when they presented the first advances in the project. During the meeting, they analyzed the letters’ metadata, such as the date, recipient and summary of the content.

Among the next steps is the application of Optical Character Recognition (OCR) techniques to transform the images of the letters into text. Machine learning models will then be used to improve the transcriptions, correcting errors and preserving relevant linguistic and paralinguistic characteristics.

“I hope to gain more knowledge about the history and potential of AI. We’ll be using these tools a lot to face and overcome challenges. It’s the first time I’ve worked with data analysis on a large, detailed database, and I believe it will contribute a lot to my academic journey,” says Nicole.

The students’ interest in the subject began back in November 2025, when João Candido presented the work of the Portinari Project at a lecture at IMPA Tech. Using some of the artist’s most iconic paintings, the students got to know part of his collection of more than 5,400 works.

Among them are Retirantes (1944), Café (1935), Criança Morta (1944) and the monumental Guerra e Paz (1952-1956), made up of two large panels commissioned by the UN (United Nations Organization).

Now, students can also get to know the artist through his epistolary output. Thanks to the preservation and digitization work carried out by the Portinari Project, the team can apply mathematical and linguistic tools and AI platforms to extract language patterns, model contexts and create vector representations capable of identifying styles and structural relationships in the texts.

With this, the project has the potential to become a relevant tool in studies on language evolution, synchronic and diachronic linguistic variations and verification of textual authorship in different contexts.