GResearch group · since 2019

GASPLN
Sentiment Analysis and Natural Language Processing

Natural language processing research focused on Portuguese — from the sentiment expressed in a text to irony, hate speech, and the language models that underpin these tasks.

Federal University of Pelotas (UFPel) Technology Development Centre (CDTec) Pelotas, Brazil

Overview

The research group on Sentiment Analysis and Natural Language Processing (GASPLN) was founded in 2019 by Professor Larissa A. de Freitas and is based at the Technology Development Centre (CDTec) at the Federal University of Pelotas (UFPel).

The work takes place within the Computer Science and Computer Engineering programmes at UFPel — at undergraduate level, through research scholarships and final-year projects, and at graduate level, through master's and doctoral research. Institutionally, the group forms part of the artificial intelligence effort brought together in the university's Hub for Innovation in Artificial Intelligence.

Shared tasks at IberLEF

The group authors the overview papers for shared tasks at the Iberian Languages Evaluation Forum, published in the journal Procesamiento del Lenguaje Natural — the paper that describes the task, the data and the results of each edition.

IDPT 2021 · Irony Detection ABSAPT 2022 · Aspect-Based Sentiment ABSAPT 2024 ASQP-PT 2025 · Sentiment Quadruples

Research lines

Sentiment Analysis

  • Sentiment analysis framework for Portuguese
  • Catalogue of sentiment analysis datasets

Sarcasm and Irony Detection

  • Sarcasm and irony detection framework for Portuguese
  • Catalogue of sarcasm and irony datasets

Hate Speech Detection

  • Hate speech detection framework for Portuguese
  • Catalogue of hate speech datasets

Data Augmentation

  • Data augmentation framework for Portuguese

Natural Language Understanding and Analysis

  • Analysis of the BERTimbau model fine-tuned for named entity recognition
  • Tool for identifying and classifying phonological errors in primary school children's writing
  • Recognition of handwritten texts by primary school children
  • Educational system supporting ENEM-style essay writing

Natural Language Modelling

  • Word embeddings to enrich a domain ontology

Researchers

Larissa Astrogildo de Freitas Founder of the group, in 2019. Sentiment analysis, irony and hate speech in Portuguese. Institutional profile Ulisses Brisolara Corrêa Aspect-based sentiment analysis, language models and the evaluation of LLMs for Portuguese. Institutional profile Brenda Salenave Santana NLP for social media, disinformation and its co-occurrence with hate speech. Institutional profile

Students and supervision

5Undergraduate research
2Final-year projects
9Master's
6Doctoral
67Completed supervisions

Current students, at undergraduate and graduate level in Computing at UFPel. The 67 completed supervisions since 2019 combine undergraduate research, final-year projects and master's degrees — the same person may appear at more than one level over the course of their studies.

Undergraduate research

  • Luciani Aquino Queiroz
  • Henrique Lopes
  • Antônio Paulo Martins Ferri
  • Thiago Kutter Neitzke
  • Luiz Felipe Pereira Bandeira

Final-year projects

  • Caio Garcez Ribeiro
  • Thomas Viatroski

Master's

  • Felix Silva
  • Andrew Campos
  • Arthur Alves Cerveira
  • Emerson Philippe Lopes
  • Guilherme Dallmann Lima
  • Iana Costa
  • João Pedro Thurow Tomaszenski
  • Lucas dos Santos Pereira
  • Henry Soares Pereira

Doctoral

  • Ivanilton Andrade
  • César Sperb
  • Carlos Augusto Calage Silveira Júnior
  • Gabriel Almeida Gomes
  • Marilia Rosa
  • Murilo Cunha

Recent publications

  • JabuticaBERT: Modern Portuguese Encoders from Scratch with RTD and Long-Context Training
    Porto, T.; Gomes, G.; Bender, A.; Corrêa, U. B.; Freitas, L. A.; Cruz, W.; Amadeus, M. · PROPOR, 2026
  • ENADE-QA: A Brazilian Portuguese Dataset for Benchmarking Large Language Models on Higher Education Exams
    Silveira, M.; Lima, G. D.; Freitas, L. A.; Corrêa, U. B. · BRACIS, 2026
  • Structured Sentiment Analysis in Brazilian Portuguese: An Exploratory Study Using BERTimbau
    Campos, A. B.; Corrêa, U. B.; Freitas, L. A. · PROPOR, 2026
  • LARI Dataset: A Native Portuguese Question Answering Dataset from Brasileiras em PLN
    Junqueira, J. R.; Freitas, L. A.; Corrêa, U. B. · PROPOR, 2026
  • Mitigating MINA-BR Unbalanced Data using Uncensored LLM for Hate Speech Detection in Brazilian Portuguese
    Campos, A.; Freitas, L. A.; Corrêa, U. B. · ENIAC, 2026
  • Avaliação End-to-End de um Sistema RAG para Documentos Hospitalares em Português
    Cunha, M. V.; Silveira, M.; Sperb, C. B.; Freitas, L. A.; Corrêa, U. B. · PROPOR, 2026
  • Overview of ASQP-PT at IberLEF 2025: Aspect-Sentiment Quadruple Prediction in Portuguese
    Lopes, E. P.; Gomes, G.; Bender, A. T.; Araújo, R. M.; Freitas, L. A.; Corrêa, U. · Procesamiento del Lenguaje Natural, v. 75, 2025
  • Small vs. Large Language Models: A Comparative Study on Multiple-Choice Question Answering in Portuguese
    Silveira, M. R.; Lima, G. D.; Júnior, C. A. C. S.; Freitas, A. L.; Corrêa, U. B. · WebMedia, 2025
  • Aya in Action: An Investigation of its Abilities in Aspect-Based Sentiment Analysis, Hate Speech Detection, Irony Detection and Question-Answering
    Junqueira, J. R.; Lopes, E. P.; Freitas, L. A.; Corrêa, U. · AMCIS, 2025
  • Discursos de Ódio e Notícias Falsas: estudo preliminar da coocorrência entre esses fenômenos em textos em português
    Parafino, C.; Santana, B. S.; Tomaszewski, J. P.; Freitas, L. A. · ERAMIA, 2025

A curated selection from 2026 and 2025, out of the group's 47 publications since 2019. Titles are kept as published.