Relatamos, neste artigo, o processo de anotação do icônico livro O Pequeno Príncipe segundo o modelo internacional Universal Dependencies, como parte de um esforço de expansão dos dados anotados atualmente disponíveis para o português do Brasil. Em especial, apresentamos o protocolo de anotação e as decisões linguísticas envolvidas, além de uma análise quantitativa do córpus.

Este trabalho investiga métodos simbólicos para a detecção de emoções em textos em português, considerando múltiplos córpus, domínios e diferentes configurações de pré-processamento. Os resultados mostram grande variação no desempenho absoluto entre domínios, mas estabilidade no desempenho relativo entre os métodos, evidenciando a influência das propriedades do córpus e o gradiente entre complexidade e interpretabilidade. A inclusão da classe neutra tende a degradar o desempenho ao aumentar a ambiguidade e, frequentemente, o desbalanceamento entre classes, enquanto um pré-processamento mais extensivo beneficia especialmente abordagens simbólicas. A análise qualitativa indica que parte dos erros decorre de ambiguidades linguísticas, do grande espaço para subjetividade no processo de anotação e das próprias nuances emocionais, reforçando a importância de avaliações comparativas multi-domínio.

Notícias falsas são um grande problema para a sociedade. Com a Inteligência Artificial generativa, notícias falsas produzidas pela máquina têm se proliferado, tornando o cenário mais desafiador. Apesar da relevância desse problema, em línguas sub-representadas como o Português, as pesquisas que buscam diferenciar notícias falsas de humanos e de máquinas são incipientes. Buscando preencher essa lacuna, este artigo explora os corpora Fake.br e FakeTrueBR expandidos com notícias falsas geradas automaticamente, caracterizando lexical e sintaticamente as notícias falsas produzidas por humanos e por máquina. Os resultados mostram que textos gerados por máquina apresentam palavras significativamente mais longas, maior uso de modificadores adjetivais e menor diversidade sintática, apesar de utilizarem mais regras sintáticas por sentença. Em contrapartida, textos humanos exibem maior variabilidade estilística em todas as dimensões analisadas.

This Ph.D. dissertation advances the state-of-the-art in Natural Language Processing (NLP) for Portuguese by proposing new and innovative data resources and explainable methods for hate speech detection and automated fact-checking. The thesis introduces several benchmark datasets for Brazilian Portuguese, HateBR, HateBRXplain, HateBRMoralXplain, MFTCXplain, MOL, and FactNews, which have been widely adopted by the research community and address critical gaps in the availability of high-quality annotated resources for Portuguese. In addition, this dissertation proposes novel post-hoc and self-explaining NLP methods: Sentence-Level Factual Reasoning (SELFAR), Social Stereotype Analysis (SSA), Contextual Bag-of-Words with Interpretable Input and Feature Optimization (B+M), Supervised Rational Attention (SRA), and Supervised Moral Rational Attention (SMRA). Across multiple tasks and datasets in Portuguese, these methods outperform baselines while improving interpretability and robustness, demonstrating that explainability and performance can be jointly optimized. Finally, this thesis has achieved significant national and international impact, being cited by leading universities and research institutes worldwide and fostering new M.Sc. and Ph.D. research projects in Brazil. Its scientific and social contributions have also been recognized with multiple prestigious national and international awards, including the Google LARA, the Maria Carolina Monard Best Thesis Award in Artificial Intelligence, the Trevisan Prize for Students “AI for Good” from Bocconi University for rigorous computer science research in AI with social impact, and the Diversity and Inclusion Award from the Association for Computational Linguistics (ACL). Lastly, this thesis has received two nominations for the Brazilian Computer Society Thesis Awards in Computer Science, and in Multimedia, Hypermedia, and Web.

Enhanced Universal Dependencies (EUD) provide a more informative syntactic representation than Basic Universal Dependencies by relaxing tree constraints to allow for graph structures. While conversion rules from basic to enhanced relations have been established for Portuguese, they were previously evaluated only on journalistic text using gold-standard basic syntactic trees. This paper evaluates the robustness of these rules in diverse scenarios (“in the wild”), encompassing other text genres and domains, as well as realistic parsing pipelines that rely on automatically generated basic syntax. Our results demonstrate that Portuguese-specific rules consistently outperform universal rules. However, the reliance on automatic basic syntax significantly impacts performance. This degradation is particularly severe when the domain of the input text differs from the training data of the basic parser. We also provide a detailed error analysis, identifying specific difficult linguistic phenomena and scenarios.

This paper reports on an effort to recover the classical morphosyntactically annotated corpus MacMorpho and realign it with the current version of the Universal Dependencies framework. We introduce a knowledge-rich approach grounded in a syntactic parser and on a specially designed tagset compatibility strategy in order to generate a “silver-standard” resource: the MacMorpho-UD-2.17. We evaluate this resource through multiple complementary methods, providing evidence for the quality of both our approach and the resulting annotation.

The spread of online misinformation has made fake news detection an essential tool for mitigating its negative impact, but many studies often disregard the temporal information, and existing datasets become outdated as news evolve. Some modern solutions using Retrieval-Augmented Generation (RAG) can solve the problem of unseen news events by providing context to the models. However, there are no studies evaluating the feasibility of web searches to attain context to decide whether a news article is true or not. This work aims to address this gap by conducting a comparative study between RAG-based solutions, traditional fake news classification methods, and deep learning-based methods. The results show that although RAG is a modern and promising technique, it cannot outperform techniques already adopted in the literature.

Enhanced Universal Dependencies (EUD) serve as a crucial link between syntax and semantics. Beyond basic syntactic dependencies, EUD provides valuable refined logical connections for downstream tasks such as semantic role labeling, coreference resolution, information extraction, and question answering. The original EUD framework defines six types of relationships, but this paper introduces an extension designed to address subject propagation in pro-drop languages. This “Extended EUD” proposal increases the number of relationships that may be annotated in sentences, improving linguistic representation. Additionally, we report our experiments on a corpus of Portuguese (a pro-drop language), which we make publicly available to the research community.

This paper presents strategies to revise an automatically annotated corpus according to the Universal Dependencies framework and discusses the learned lessons, mainly regarding the annotators’ behavior. The revision strategies are not relying on examples from any specific language and, because they are languageindependent, can be adopted in any language and corpus annotation initiative.

Abstract Meaning Representation (AMR) is a meaning representation (MR) designed to abstract away from syntax, allowing syntactically different sentences to share the same AMR graph. Unlike other MRs, existing AMR corpora typically link one AMR graph to a single reference. This paper investigates the value of paraphrase generation in low-resource AMR-to-Text generation by testing various paraphrase generation strategies and evaluating their impact. The findings show that paraphrase generation significantly outperforms the baseline and traditional data augmentation methods, even with fewer training instances. Human evaluations indicate that this strategy often produces syntactic-based paraphrases and can exceed the performance of previous approaches. Additionally, the paper releases a paraphrase-extended version of the AMR corpus.

Most existing fact-checking systems are unable to explain their decisions by providing relevant rationales (justifications) for their predictions. It highlights a lack of transparency that poses significant risks, such as the prevalence of unexpected biases, which may increase political polarization due to limitations in impartiality. To address this critical gap, we introduce SEntence-Level FActual Reasoning (SELFAR), aimed at improving explainable fact-checking. SELFAR relies on fact extraction and verification by predicting the news source reliability and factuality (veracity) of news articles or claims at the sentence level, generating post-hoc explanations using SHAP/LIME and zero-shot prompts. Our experiments show that unreliable news stories predominantly consist of subjective statements, in contrast to reliable ones. Consequently, predicting unreliable news articles at the sentence level by analyzing impartiality and subjectivity is a promising approach for fact extraction and improving explainable fact-checking. Furthermore, LIME outperforms SHAP in explaining predictions on reliability. Additionally, while zero-shot prompts provide highly readable explanations and achieve an accuracy of 0.71 in predicting factuality, their tendency to hallucinate remains a challenge. Lastly, this paper also presents the first study on explainable fact-checking in the Portuguese language.

One of the challenges of aspect-based sentiment analysis is the implicit mention of aspects. These are more difficult to identify and may require world knowledge to do so. In this work, we evaluate frequency-based, hybrid, and machine learning methods, including the use of the pre-trained BERT language model, in the task of extracting aspect terms in opinionated texts in Portuguese, emphasizing the analysis of implicit aspects. Besides the comparative evaluation of methods, the differential of this work lies in the analysis’s novelty using a typology of implicit aspects that shows the knowledge needed to identify each implicit aspect term, thus allowing a mapping of the strengths and weaknesses of each method.

This paper describes the submission by the NILC Computational Linguistics research group of the University of São Paulo/Brazil to the RDF-to-Text task for English at the WebNLG+ challenge. The success of the current pretrained models like BERT or GPT-2 in text-to-text generation tasks is well-known, however, its application/success on data-totext generation has not been well-studied and proven. This way, we explore how good a pretrained model, in particular BART, performs on the data-to-text generation task. The results obtained were worse than the baseline and other systems in almost all automatic measures. However, the human evaluation shows better results for our system. Besides, results suggest that BART may generate paraphrases of reference texts.