Vinicius Ruela Pereira Borges
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Vinicius Ruela Pereira Borges has a total of 77 co-authors. Here are the top co-authors by number of shared publications:
- Maristela Holanda
Universidade de Brasília
4 shared publications - Mayara Chew Marinho
Departamento de Ciência da Computac ¸ão Universidade de Brasília (UnB) Brasília -DF Brazil
4 shared publications - Bernardo Romão
Faculty of Health Sciences, Department of Nutrition, University of Brasília, Brasília 70910-900, Brazil
3 shared publications Celia A. Zorzo Barcelos
3 shared publications- Maria Cristina Ferreira de Oliveira
University of São Paulo
3 shared publications - Renata Puppin Zandonadi
Department of Nutrition, Faculty of Health Sciences, University of Brasília, Campus Universitario Darcy Ribeiro, Brasília 70910-900, Brazil
3 shared publications
Latest Publications
- 3 October 2026
- Published by Springer Nature
- p. 592-607
Evaluating Location-Aware RAG Approaches for Question Answering over Public Health–Related Content in Brazilian Official Gazettes
- 3 October 2026
- Published by Springer Nature
- p. 592-607
Evaluating Location-Aware RAG Approaches for Question Answering over Public Health–Related Content in Brazilian Official Gazettes
- 16 November 2025
- Published by Springer Nature
- p. 555-567
Evaluating Cross-Lingual Semantic Textual Similarity Using BERT-Based Sentence Embeddings
- 16 November 2025
- Published by Springer Nature
- p. 555-567
Evaluating Cross-Lingual Semantic Textual Similarity Using BERT-Based Sentence Embeddings
- Open Access
- 20 July 2025
- Published by Sociedade Brasileira de Computação
- p. 203-214
Automatic Classification of Access Levels in Documents from the Brazilian Electronic Information System
The Brazilian Electronic Information System (SEI) is used to manage documents and administrative processes in public institutions. Users often need to upload external documents and manually assign an access level: public, restricted, or confidential. To reduce human errors in this process, we propose an automatic classification approach. A labeled corpus was constructed using public SEI’s documents and artificial samples generated from SEI’s form templates filled with fictional content using ChatGPT. We trained and evaluated four classification models: SVM, LSTM, BiLSTM, and BERT. Experiments were performed by means of Stratified K-Fold cross-validation and the results showed that SVM and BiLSTM performed best, each achieving a macro F1-score of 0.96.
- 20 July 2025
- Published by Sociedade Brasileira de Computação
- p. 203-214
Open AccessAutomatic Classification of Access Levels in Documents from the Brazilian Electronic Information System
The Brazilian Electronic Information System (SEI) is used to manage documents and administrative processes in public institutions. Users often need to upload external documents and manually assign an access level: public, restricted, or confidential. To reduce human errors in this process, we propose an automatic classification approach. A labeled corpus was constructed using public SEI’s documents and artificial samples generated from SEI’s form templates filled with fictional content using ChatGPT. We trained and evaluated four classification models: SVM, LSTM, BiLSTM, and BERT. Experiments were performed by means of Stratified K-Fold cross-validation and the results showed that SVM and BiLSTM performed best, each achieving a macro F1-score of 0.96.
- Open Access
- 19 May 2025
- Published by Sociedade Brasileira de Computação
- p. 778-787
Could you tell me the process ID? Structuring Text Documents from the Brazilian Electronic Information System Using a Named Entity Recognition Approach
Context: In the context of valuing open data, transparency, and efficiency in public services, there is a growing demand for studies to improve government systems. Some public agencies use the Brazilian Electronic Information System (Sistema Eletrônico de Informações - SEI), a procedural management system that centralizes electronic processes and promotes administrative efficiency. Problem: Although SEI has contributed to advancements in public administration, significant challenges remain in information search and retrieval due to the inefficient keyword-based approach currently available. These difficulties are enhanced by the high amount of documents generated daily, which are written in natural language and present variability in categories, writing styles, and structures. As a result, searching for relevant documents in SEI is time-consuming, leading users to create unnecessary processes and inconsistent resolutions when compared to previously completed processes. Solution: A Natural Language Processing (NLP) pipeline was proposed to extract information from SEI documents using Named Entity Recognition (NER) models. IS Theory: Organizational Information Processing. Method: This research adopts a descriptive approach. Public SEI documents were collected using a web crawler, and trained annotators built a corpus to enable the training of state-of-the-art NER models. The models’ performances were compared and quantitatively analyzed. Summary of Results: A Brazilian Portuguese labeled corpus of SEI for NER was curated and validated, leading to an NLP pipeline for information extraction. Contributions and Impact in the IS area: This research provides a baseline for structuring data from Electronic Information Systems, enabling more effective strategies for search and retrieval tasks.
- 19 May 2025
- Published by Sociedade Brasileira de Computação
- p. 778-787
Open AccessCould you tell me the process ID? Structuring Text Documents from the Brazilian Electronic Information System Using a Named Entity Recognition Approach
Context: In the context of valuing open data, transparency, and efficiency in public services, there is a growing demand for studies to improve government systems. Some public agencies use the Brazilian Electronic Information System (Sistema Eletrônico de Informações - SEI), a procedural management system that centralizes electronic processes and promotes administrative efficiency. Problem: Although SEI has contributed to advancements in public administration, significant challenges remain in information search and retrieval due to the inefficient keyword-based approach currently available. These difficulties are enhanced by the high amount of documents generated daily, which are written in natural language and present variability in categories, writing styles, and structures. As a result, searching for relevant documents in SEI is time-consuming, leading users to create unnecessary processes and inconsistent resolutions when compared to previously completed processes. Solution: A Natural Language Processing (NLP) pipeline was proposed to extract information from SEI documents using Named Entity Recognition (NER) models. IS Theory: Organizational Information Processing. Method: This research adopts a descriptive approach. Public SEI documents were collected using a web crawler, and trained annotators built a corpus to enable the training of state-of-the-art NER models. The models’ performances were compared and quantitatively analyzed. Summary of Results: A Brazilian Portuguese labeled corpus of SEI for NER was curated and validated, leading to an NLP pipeline for information extraction. Contributions and Impact in the IS area: This research provides a baseline for structuring data from Electronic Information Systems, enabling more effective strategies for search and retrieval tasks.
- Open Access
- 19 February 2025
- Published by Sociedade Brasileira de Computação in Computação Brasil
- No. 53,p. 24-28
Maratona de Programação: Rumo ao Futuro
Programação competitiva engaja e motiva alunos a desenvolverem suas habilidades de resolução de problemas, tornando-os melhor preparados para o mercado de trabalho. A Maratona SBC de Programação é o principal evento do tipo no Brasil, envolvendo milhares de alunos de centenas de instituições de ensino e fomentando a diversidade e o networking entre eles. Apesar de muitas enfrentarem dificuldades relacionadas à falta de recursos, infraestrutura, suporte logístico e professores especializados, o que resulta em uma diferença significativa de desempenho e representatividade entre elas, os benefícios colhidos compensam os esforços dos competidores e voluntários envolvidos.
- 19 February 2025
- Published by Sociedade Brasileira de Computação in Computação Brasil
- No. 53,p. 24-28
Open AccessMaratona de Programação: Rumo ao Futuro
Programação competitiva engaja e motiva alunos a desenvolverem suas habilidades de resolução de problemas, tornando-os melhor preparados para o mercado de trabalho. A Maratona SBC de Programação é o principal evento do tipo no Brasil, envolvendo milhares de alunos de centenas de instituições de ensino e fomentando a diversidade e o networking entre eles. Apesar de muitas enfrentarem dificuldades relacionadas à falta de recursos, infraestrutura, suporte logístico e professores especializados, o que resulta em uma diferença significativa de desempenho e representatividade entre elas, os benefícios colhidos compensam os esforços dos competidores e voluntários envolvidos.
- Open Access
- 7 January 2025
- Published by MDPI AG in Sustainability
- Vol. 17 (2) , 382
Sustainability of Key Proteins in Plant-Based Meat Analogs Production: A Worldwide Perspective
The market for plant-based analogs for meat is growing exponentially. In addition to motivations related to the search for health benefits, the consumption of such products is justified by the sustainability of their production since the use of non-renewable resources and the emission of polluting gases is lower than their animal-origin equivalents. However, little information regarding the global panorama of the sustainability of plant-based meat analogues is available, mainly due to the diffuse distribution of food matrices used across the planet. In this sense, this narrative review aimed to describe the state of the art regarding the use of resources and sustainability of the inputs used as protein sources in the manufacture of plant-based meat analogues. From the review carried out, it was possible to observe that the biggest problem in producing these plant-based alternatives lies in using inputs that are not native to the countries where the products are marketed, especially in the case of South American countries. Ingredients widely used in the production of these analogues find better cultivation conditions in the northern hemisphere, as in the case of lentils, peas and chickpeas; thus, South American markets depend on imports, reducing the sustainability of the products.
- 7 January 2025
- Published by MDPI AG in Sustainability
- Vol. 17 (2) , 382
Open AccessSustainability of Key Proteins in Plant-Based Meat Analogs Production: A Worldwide Perspective
The market for plant-based analogs for meat is growing exponentially. In addition to motivations related to the search for health benefits, the consumption of such products is justified by the sustainability of their production since the use of non-renewable resources and the emission of polluting gases is lower than their animal-origin equivalents. However, little information regarding the global panorama of the sustainability of plant-based meat analogues is available, mainly due to the diffuse distribution of food matrices used across the planet. In this sense, this narrative review aimed to describe the state of the art regarding the use of resources and sustainability of the inputs used as protein sources in the manufacture of plant-based meat analogues. From the review carried out, it was possible to observe that the biggest problem in producing these plant-based alternatives lies in using inputs that are not native to the countries where the products are marketed, especially in the case of South American countries. Ingredients widely used in the production of these analogues find better cultivation conditions in the northern hemisphere, as in the case of lentils, peas and chickpeas; thus, South American markets depend on imports, reducing the sustainability of the products.
- Open Access
- 17 November 2024
- Published by Sociedade Brasileira de Computação
- p. 73-84
Pre-trained Language Models for Multi-Label Text Classification of Competitive Programming Problems
This paper explores the use of pre-trained language models for classifying programming problems from online judges based on topics commonly addressed in competitive programming. State-of-the-art language models were employed as text classifiers, including Long Short-Term Memory (LSTM) and Bidirectional LSTM with pre-trained Word2Vec embeddings, Bidirectional Encoder Representations from Transformers (BERT), and Llama3.1-8B. Experiments were conducted using two different representations of the programming problems: standalone statement and statement with source code. The results showed that Llama3.1-8B achieved the best overall Macro F1-Score, outperforming the other models by a significant margin.
- 17 November 2024
- Published by Sociedade Brasileira de Computação
- p. 73-84
Open AccessPre-trained Language Models for Multi-Label Text Classification of Competitive Programming Problems
This paper explores the use of pre-trained language models for classifying programming problems from online judges based on topics commonly addressed in competitive programming. State-of-the-art language models were employed as text classifiers, including Long Short-Term Memory (LSTM) and Bidirectional LSTM with pre-trained Word2Vec embeddings, Bidirectional Encoder Representations from Transformers (BERT), and Llama3.1-8B. Experiments were conducted using two different representations of the programming problems: standalone statement and statement with source code. The results showed that Llama3.1-8B achieved the best overall Macro F1-Score, outperforming the other models by a significant margin.
- Open Access
- 4 November 2024
- Published by Sociedade Brasileira de Computação
- p. 1796-1809
Natural Language Processing Approaches for Accrediting Students on Extracurricular Activities
The undergraduate programs at Brazilian universities allow students to include extracurricular activities in their academic transcripts. The large amount of proof documents (certificates and declarations) submitted by students that are subsequently analyzed by the academic staff makes the accrediting of extracurricular activities time-consuming and prone to error. This paper describes a methodology to classify academic proof documents according to the pre-defined groups by the Universidade de Brasília regulations for extracurricular activities accreditation. Experimental results showed that TF-IDF with SVM outperformed BERT, CNN and BiLSTM with 0.94 average Macro F1-Score, though their performances' difference were not statistically significant.
- 4 November 2024
- Published by Sociedade Brasileira de Computação
- p. 1796-1809
Open AccessNatural Language Processing Approaches for Accrediting Students on Extracurricular Activities
The undergraduate programs at Brazilian universities allow students to include extracurricular activities in their academic transcripts. The large amount of proof documents (certificates and declarations) submitted by students that are subsequently analyzed by the academic staff makes the accrediting of extracurricular activities time-consuming and prone to error. This paper describes a methodology to classify academic proof documents according to the pre-defined groups by the Universidade de Brasília regulations for extracurricular activities accreditation. Experimental results showed that TF-IDF with SVM outperformed BERT, CNN and BiLSTM with 0.94 average Macro F1-Score, though their performances' difference were not statistically significant.
- Open Access
- 20 May 2024
- Published by Sociedade Brasileira de Computação
- p. 212-221
Estratégias Computacionais Baseadas em Similaridade de Textos e Visualização Exploratória para a Identificação de Inconsistências em Notas Fiscais Eletrônicas
A detecção de fraudes em notas fiscais é um desafio considerável para os órgãos de controle, dada a quantidade expressiva de documentos gerados diariamente. Buscando soluções para esse desafio, esta pesquisa propõe uma abordagem inovadora que combina rotulação automática de notas fiscais baseada em similaridade de texto e técnicas de visualização exploratória. O objetivo é auxiliar especialistas na tarefa de auditoria fiscal, trazendo equilíbrio entre automatização e autonomia por meio de um sistema de visualização web interativo denominado NFViz. O NFViz permite investigar as notas fiscais em busca de incompatibilidades que podem sinalizar possíveis fraudes, otimizando as atividades de auditoria fiscal ao reduzir o escopo de busca por notas suspeitas.
- 20 May 2024
- Published by Sociedade Brasileira de Computação
- p. 212-221
Open AccessEstratégias Computacionais Baseadas em Similaridade de Textos e Visualização Exploratória para a Identificação de Inconsistências em Notas Fiscais Eletrônicas
A detecção de fraudes em notas fiscais é um desafio considerável para os órgãos de controle, dada a quantidade expressiva de documentos gerados diariamente. Buscando soluções para esse desafio, esta pesquisa propõe uma abordagem inovadora que combina rotulação automática de notas fiscais baseada em similaridade de texto e técnicas de visualização exploratória. O objetivo é auxiliar especialistas na tarefa de auditoria fiscal, trazendo equilíbrio entre automatização e autonomia por meio de um sistema de visualização web interativo denominado NFViz. O NFViz permite investigar as notas fiscais em busca de incompatibilidades que podem sinalizar possíveis fraudes, otimizando as atividades de auditoria fiscal ao reduzir o escopo de busca por notas suspeitas.
- Open Access
- 22 April 2024
- Published by Sociedade Brasileira de Computação
- p. 102-112
Análise do Uso de Componentes do Moodle em uma Disciplina Introdutória de Programação
A primeira disciplina de programação é uma das bases para os cursos de graduação na área de computação, por exercer um papel importante no desenvolvimento da habilidade de programar. De maneira geral, essa disciplina apresenta altas taxas de reprovação tanto no Brasil quanto em outros países, o que tem gerado diversas iniciativas para a promoção de melhorias na metodologia de ensino. Na Universidade de Brasília, a disciplina inicial de programação tem utilizado o Moodle como ambiente de aprendizagem. Um dos desafios dos docentes neste ambiente é entender como os alunos utilizam os componentes criados no seu processo de aprendizagem. Neste contexto, este trabalho apresenta uma análise dos componentes criados no Moodle da disciplina inicial de programação, considerando questionários e lições, com o objetivo de entender como os alunos utilizaram esses dois componentes de aprendizagem. Os dados apresentados são dos períodos da pandemia com o ensino totalmente remoto e o primeiro semestre do retorno presencial. Os resultados mostram que em geral a realização dessas tarefas costuma ser mais frequente nas primeiras semanas da disciplina. Além disso, quando essas tarefas passam a ser optativas, na volta ao ensino presencial, a participação diminui.
- 22 April 2024
- Published by Sociedade Brasileira de Computação
- p. 102-112
Open AccessAnálise do Uso de Componentes do Moodle em uma Disciplina Introdutória de Programação
A primeira disciplina de programação é uma das bases para os cursos de graduação na área de computação, por exercer um papel importante no desenvolvimento da habilidade de programar. De maneira geral, essa disciplina apresenta altas taxas de reprovação tanto no Brasil quanto em outros países, o que tem gerado diversas iniciativas para a promoção de melhorias na metodologia de ensino. Na Universidade de Brasília, a disciplina inicial de programação tem utilizado o Moodle como ambiente de aprendizagem. Um dos desafios dos docentes neste ambiente é entender como os alunos utilizam os componentes criados no seu processo de aprendizagem. Neste contexto, este trabalho apresenta uma análise dos componentes criados no Moodle da disciplina inicial de programação, considerando questionários e lições, com o objetivo de entender como os alunos utilizaram esses dois componentes de aprendizagem. Os dados apresentados são dos períodos da pandemia com o ensino totalmente remoto e o primeiro semestre do retorno presencial. Os resultados mostram que em geral a realização dessas tarefas costuma ser mais frequente nas primeiras semanas da disciplina. Além disso, quando essas tarefas passam a ser optativas, na volta ao ensino presencial, a participação diminui.