Automatic Assessment of Argumentation Quality
ArgQ! is a research project that aims to develop models to automatically assess the quality of arguments in short messages from social networks, such as X (formerly Twitter). The project includes creating an annotated corpus and computational models for evaluating rhetorical quality in Portuguese.
About the Project
A computational approach for automatically assessing argumentation quality in the rhetorical dimension of tweets related to Brazilian politics.
Argumentation is fundamental to human communication and has been studied across multiple disciplines, including computer science. With the rise of social networks as a source of argumentative content, Natural Language Processing (NLP) has become a key tool for automatic argument evaluation.
This project contributes to Portuguese-language research by proposing and validating a computational model fine-tuned with BERTimbau to evaluate rhetorical quality in Brazilian political tweets.
Researcher
Cássio Faria da Silva
Laboratory of Linguistics and Computational Intelligence (LALIC)
Department of Computer Science, UFSCar
https://www.lalic.ufscar.br
cassiofs [at] gmail.com
Project Goal
Develop computational models for automatically evaluating the quality of arguments in short social media messages.
Methodology
Construction of an annotated corpus and training of machine learning models to assess rhetorical quality in arguments.
Annotation Team
Amanda Pontes Rassi
Jackson Wilke C. Souza
Renata Ramisch
Roger Alfredo M. R. Antunes
Supervision
Dr. Helena de Medeiros Caseli
Dr. Vânia Paula de Almeida Neris
Annotation Guidelines
Detailed instructions for corpus annotation and quality criteria.
Corpus Annotation Guidelines
The document below describes the annotation process, classification criteria, and examples used in this project to ensure consistency and accuracy. It serves as a reference for researchers interested in replicating or extending this work.
📄 Download Guidelines (PDF)Interface Prototype
Public demonstration of the system’s interface and functionality.
Interactive Prototype
Explore the public interface demonstrating how the system evaluates argumentative quality in tweets about Brazilian politics. Data is used exclusively for academic purposes.
Access PrototypeRelated Publications
This section gathers the main publications derived from this doctoral research. Each article describes the methodologies, experiments, and contributions that support the development of the ArgQ! model. Access the works below for detailed information, results, and future research perspectives.
Quality of Argumentation in Political Tweets: What It Is and How to Measure It
Journal of Language Studies (RELIN)
Published in volume 29(4) of RELIN, this article presents the full process of corpus creation and
annotation, including guideline definition and inter-annotator agreement analysis. It provides
theoretical grounding for assessing argumentation quality and includes examples from Twitter data.
Reference:
Cássio Faria da Silva et al. (2021). Quality of argumentation in political tweets: what it is and
how to measure it. Revista de Estudos da Linguagem, 29(4), 2537–2586.
DOI:
10.17851/2237-2083.29.4.2537-2586
Argument Quality Assessment in Brazilian Political Tweets
Revista Linguamática
This 2023 article describes computational experiments using feature-based and neural models
(BERTimbau and RoBERTaTwitterBR) to automatically evaluate rhetorical quality in Portuguese tweets.
It details annotation, taxonomy, evaluation metrics, and performance results.
Reference:
SILVA, Cássio Faria da; NERIS, Vânia Paula de Almeida; CASELI, Helena de Medeiros (2023).
Argument Quality Assessment in Brazilian Political Tweets. Linguamática, 15(1), 102–127.
DOI:
10.21814/lm.15.1.387
Doctoral Thesis
Department of Computer Science — UFSCar
This thesis consolidates the research results, including the annotated corpus, proposed guidelines,
and computational models. It discusses theoretical background, experiments, and implications for
computational argumentation research in Portuguese.
Reference:
SILVA, Cássio Faria da (2023). Computational Approach for Automatic Assessment of Argumentation
Quality in the Rhetorical Dimension of Tweets Related to Brazilian Politics. Doctoral Thesis
(Computer Science). Federal University of São Carlos (UFSCar).
Repository:
https://repositorio.ufscar.br/handle/ufscar/18253
Contact
Get in touch with the research team or explore institutional links below.
Research Conducted At
Laboratory of Linguistics and Computational Intelligence (LALIC)
Department of Computer Science — UFSCar
Contact the Researcher
Cássio Faria da Silva
Email: cassiofs [at] gmail.com