FORMAS aims to study, analyze and evaluate semantic-based approaches. Our main research areas are based on five major pillars on semantic-based area: Methods, Ontology, Information Extraction, Interoperability and Big Data. We deeply analyze each one of these approaches focusing on obtaining high levels of semantic and pragmatic comprehension. Visite our website for more informations.

• Attention in Transformer investigates how attention weights in a Transformer model behave when anchored to explicit governor-dependent relations extracted from UD-annotated corpora.

  • Syntactic Analysis in Transformers through Attention Heads @STIL2025 - 2nd Best Paper Award
  • Grammatical Representations in Transformer Attention: A Multidomain Study of Portuguese via UD @BRACIS2026

• CSIS method interoperates Syntactic, Semantic and Pragmatic into University Surveillance models providing image captioning for operators and systems.

  • Towards a SoIS model for University Surveillance @ SESOS2025
  • Supporting Pragmatic Interoperability: An LLM Based Process to Analyze Distributed Intentionality (i*) Models @ MODRE2025
  • CSIS Ecosystem: Impacts of Pragmatic Detection on a Campus Surveillance Case Study @SBSI2026

• DIGGER method uses LLMs to provide a QA for the CDC legal documents.

  • Brazilian Consumer Protection Code: a methodology for a dataset to Question-Answer (QA) Models @PADAWAN2024
  • Towards a Corpus Methodology for LLMs in the Legal Domain @STIL2025

• DptOIE method extract triples from Universal Dependencies (UD) format.

  • DptOIE: a Portuguese open information extraction based on dependency analysis. @AIR JOURNAL
  • [DPToie-Python]

• PortNOIE is a new version of DPTOIE-Neural.

  • PortNOIE: A Neural Framework for Open Information Extraction for the Portuguese Language @PROPOR2024
  • Extração de Informação Aberta com LLM para a Língua Portuguesa @LINGUAMATICA
  • Exploring Open Information Extraction for Portuguese Using Large Language Models @PROPOR2024
  • Scaling and Adapting Large Language Models for Portuguese Open Information Extraction: A Comparative Study of Fine-Tuning and LoRA @BRACIS2024

• PTOIE-Flair is a pt-br OpenIE model.

  • Exploring Open Information Extraction for the Portuguese language: An integrated monolithic approach in Cloud environment @PROPOR2024

• PragmaticOIE method uses a rule-based approach to extract facts in Portuguese in a first pragmatic level.

  • PragmaticOIE: a pragmatic open information extraction for Portuguese language @NLP JOURNAL

• ImageCaptioningPT methods to generate image captioning in the Portuguese language.

  • Towards Image Captioning for the Portuguese Language: Evaluation on a Translated Dataset @ICEIS2023
  • A bilingual analysis of multi-head attention mechanism for image captioning based on morphosyntactic information @JBCS2025
  • Analysis of Machine Translators on Sentences Generated by Portuguese Image Captioning Models @PROPOR 2026

• ALiBWeb is a Web system to map Brazilian dialectology areas.

  • ALiBWeb: estado da arte e perspectivas futuras @WORKINGPAPER