Marco Antonio Sobrevilla Cabezudo
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
Marco Antonio Sobrevilla Cabezudo has a total of 45 co-authors. Here are the top co-authors by number of shared publications:
- Thiago Alexandre Salgueiro Pardo
University of São Paulo
6 shared publications - Marcio Lima Inácio
University of Coimbra
4 shared publications - Ariani Di Felippo
UFSCar
2 shared publications - Rafael Torres Anchiêta
Federal Institute of Piaui
2 shared publications - Renata Ramisch
Redação Nota 1000, Brasil
2 shared publications Alessandro Bokan
1 shared publication
Latest Publications
- Open Access
- 1 January 2024
- Published by Association for Computational Linguistics (ACL)
- p. 663-675
Investigating Paraphrase Generation as a Data Augmentation Strategy for Low-Resource AMR-to-Text Generation
Marco Antonio Sobrevilla Cabezudo, Marcio Lima Inacio, Thiago Alexandre Salgueiro Pardo. Proceedings of the 17th International Natural Language Generation Conference. 2024.
- 1 January 2024
- Published by Association for Computational Linguistics (ACL)
- p. 663-675
Open AccessInvestigating Paraphrase Generation as a Data Augmentation Strategy for Low-Resource AMR-to-Text Generation
Marco Antonio Sobrevilla Cabezudo, Marcio Lima Inacio, Thiago Alexandre Salgueiro Pardo. Proceedings of the 17th International Natural Language Generation Conference. 2024.
- Open Access
- 1 January 2023
- Published by FapUNIFESP (SciELO) in DELTA: Documentação de Estudos em Lingüística Teórica e Aplicada
- Vol. 39 (3)
The AMR-PT corpus and the semantic annotation of challenging sentences from journalistic and opinion texts
One of the most popular semantic representation languages in Natural Language Processing (NLP) is Abstract Meaning Representation (AMR). This formalism encodes the meaning of single sentences in directed rooted graphs. For English, there is a large annotated corpus that provides qualitative and reusable data for building or improving existing NLP methods and applications. For building AMR corpora for non-English languages, including Brazilian Portuguese, automatic and manual strategies have been conducted. The automatic annotation methods are essentially based on the cross-linguistic alignment of parallel corpora and the inheritance of the AMR annotation. The manual strategies focus on adapting the AMR English guidelines to a target language. Both annotation strategies have to deal with some phenomena that are challenging. This paper explores in detail some characteristics of Portuguese for which the AMR model had to be adapted and introduces two annotated corpora: AMRNews, a corpus of 870 annotated sentences from journalistic texts, and OpiSums-PT-AMR, comprising 404 opinionated sentences in AMR.
- 1 January 2023
- Published by FapUNIFESP (SciELO) in DELTA: Documentação de Estudos em Lingüística Teórica e Aplicada
- Vol. 39 (3)
Open AccessThe AMR-PT corpus and the semantic annotation of challenging sentences from journalistic and opinion texts
One of the most popular semantic representation languages in Natural Language Processing (NLP) is Abstract Meaning Representation (AMR). This formalism encodes the meaning of single sentences in directed rooted graphs. For English, there is a large annotated corpus that provides qualitative and reusable data for building or improving existing NLP methods and applications. For building AMR corpora for non-English languages, including Brazilian Portuguese, automatic and manual strategies have been conducted. The automatic annotation methods are essentially based on the cross-linguistic alignment of parallel corpora and the inheritance of the AMR annotation. The manual strategies focus on adapting the AMR English guidelines to a target language. Both annotation strategies have to deal with some phenomena that are challenging. This paper explores in detail some characteristics of Portuguese for which the AMR model had to be adapted and introduces two annotated corpora: AMRNews, a corpus of 870 annotated sentences from journalistic texts, and OpiSums-PT-AMR, comprising 404 opinionated sentences in AMR.
- Top Cited PapersOpen Access
- 1 January 2023
- Published by Association for Computational Linguistics (ACL)
- p. 947-959
The Dangers of trusting Stochastic Parrots: Faithfulness and Trust in Open-domain Conversational Question Answering
Sabrina Chiesurin, Dimitris Dimakopoulos, Marco Antonio Sobrevilla Cabezudo, Arash Eshghi, Ioannis Papaioannou, Verena Rieser, Ioannis Konstas. Findings of the Association for Computational Linguistics: ACL 2023. 2023.
- 1 January 2023
- Published by Association for Computational Linguistics (ACL)
- p. 947-959
Top Cited PapersOpen AccessThe Dangers of trusting Stochastic Parrots: Faithfulness and Trust in Open-domain Conversational Question Answering
Sabrina Chiesurin, Dimitris Dimakopoulos, Marco Antonio Sobrevilla Cabezudo, Arash Eshghi, Ioannis Papaioannou, Verena Rieser, Ioannis Konstas. Findings of the Association for Computational Linguistics: ACL 2023. 2023.
- Open Access
- 11 November 2022
- Published by Springer Nature in Research Square
Comparison of Cross-lingual Strategies for AMR-to-Brazilian Portuguese Generation
This work presents a study of different strategies for tackling low-resource AMR-to-text generation for Brazilian Portuguese on three approaches. In particular, we explore the use of English AMR as an interlingua and transfer learning (TL). The results suggest that using AMR as an interlingua can be a strong baseline. However, we need to consider the bilingual dictionary used in the concept translation, as it can harm text generation. On the other hand, TL seems to be a promising strategy for generating accurate outputs, but its contribution is not significant. Finally, we present a transformer-based model with the best performance, surpassing all baselines.
- 11 November 2022
- Published by Springer Nature in Research Square
Open AccessComparison of Cross-lingual Strategies for AMR-to-Brazilian Portuguese Generation
This work presents a study of different strategies for tackling low-resource AMR-to-text generation for Brazilian Portuguese on three approaches. In particular, we explore the use of English AMR as an interlingua and transfer learning (TL). The results suggest that using AMR as an interlingua can be a strong baseline. However, we need to consider the bilingual dictionary used in the concept translation, as it can harm text generation. On the other hand, TL seems to be a promising strategy for generating accurate outputs, but its contribution is not significant. Finally, we present a transformer-based model with the best performance, surpassing all baselines.
- Open Access
- 30 August 2022
- Published by FapUNIFESP (SciELO) in SciELO Preprints
The AMR-PT corpus and the semantic annotation of challenging sentences from journalistic and opinion texts
One of the most popular semantic representation languages in Natural Language Processing (NLP) is Abstract Meaning Representation (AMR). This formalism encodes the meaning of single sentences in directed rooted graphs. For English, there is a large annotated corpus that provides qualitative and reusable data for building or improving existing NLP methods and applications. For building AMR corpora for non-English languages, including Brazilian Portuguese, automatic and manual strategies have been conducted. The automatic annotation methods are essentially based on the cross-linguistic alignment of parallel corpora and the inheritance of the AMR annotation. The manual strategies focus on adapting the AMR English guidelines to a target language. Both annotation strategies have to deal with some phenomena that are challenging. This paper explores in detail some characteristics of Portuguese for which the AMR model had to be adapted and introduces two annotated corpora: AMRNews, a corpus of 870 annotated sentences from journalistic texts, and OpiSums-PT-AMR, comprising 404 opinionated sentences in AMR.
- 30 August 2022
- Published by FapUNIFESP (SciELO) in SciELO Preprints
Open AccessThe AMR-PT corpus and the semantic annotation of challenging sentences from journalistic and opinion texts
One of the most popular semantic representation languages in Natural Language Processing (NLP) is Abstract Meaning Representation (AMR). This formalism encodes the meaning of single sentences in directed rooted graphs. For English, there is a large annotated corpus that provides qualitative and reusable data for building or improving existing NLP methods and applications. For building AMR corpora for non-English languages, including Brazilian Portuguese, automatic and manual strategies have been conducted. The automatic annotation methods are essentially based on the cross-linguistic alignment of parallel corpora and the inheritance of the AMR annotation. The manual strategies focus on adapting the AMR English guidelines to a target language. Both annotation strategies have to deal with some phenomena that are challenging. This paper explores in detail some characteristics of Portuguese for which the AMR model had to be adapted and introduces two annotated corpora: AMRNews, a corpus of 870 annotated sentences from journalistic texts, and OpiSums-PT-AMR, comprising 404 opinionated sentences in AMR.
- Open Access
Efficient strategies for hierarchical text classification: External knowledge and auxiliary tasks
In hierarchical text classification, we perform a sequence of inference steps to predict the category of a document from top to bottom of a given class taxonomy. Most of the studies have focused on developing novels neural network architectures to deal with the hierarchical structure, but we prefer to look for efficient ways to strengthen a baseline model. We first define the task as a sequence-to-sequence problem. Afterwards, we propose an auxiliary synthetic task of bottom-up-classification. Then, from external dictionaries, we retrieve textual definitions for the classes of all the hierarchy's layers, and map them into the word vector space. We use the class-definition embeddings as an additional input to condition the prediction of the next layer and in an adapted beam search. Whereas the modified search did not provide large gains, the combination of the auxiliary task and the additional input of class-definitions significantly enhance the classification accuracy. With our efficient approaches, we outperform previous studies, using a drastically reduced number of parameters, in two well-known English datasets.
Open AccessEfficient strategies for hierarchical text classification: External knowledge and auxiliary tasks
In hierarchical text classification, we perform a sequence of inference steps to predict the category of a document from top to bottom of a given class taxonomy. Most of the studies have focused on developing novels neural network architectures to deal with the hierarchical structure, but we prefer to look for efficient ways to strengthen a baseline model. We first define the task as a sequence-to-sequence problem. Afterwards, we propose an auxiliary synthetic task of bottom-up-classification. Then, from external dictionaries, we retrieve textual definitions for the classes of all the hierarchy's layers, and map them into the word vector space. We use the class-definition embeddings as an additional input to condition the prediction of the next layer and in an adapted beam search. Whereas the modified search did not provide large gains, the combination of the auxiliary task and the additional input of class-definitions significantly enhance the classification accuracy. With our efficient approaches, we outperform previous studies, using a drastically reduced number of parameters, in two well-known English datasets.
- 24 February 2020
- Published by Springer Nature
- p. 346-356
Natural Language Inference for Portuguese Using BERT and Multilingual Information
- 24 February 2020
- Published by Springer Nature
- p. 346-356
Natural Language Inference for Portuguese Using BERT and Multilingual Information
- 7 October 2019
- Published by Instituto Politecnico Nacional/Centro de Investigacion en Computacion in Computación y Sistemas
- Vol. 23 (3) , 731-740
An Exploratory Study of the Use of Senses, Syntax and Cross-Linguistic Information for Subjectivity Detection in Spanish
An Exploratory Study of the Use of Senses, Syntax and Cross-Linguistic Information for Subjectivity Detection in Spanish
- 7 October 2019
- Published by Instituto Politecnico Nacional/Centro de Investigacion en Computacion in Computación y Sistemas
- Vol. 23 (3) , 731-740
An Exploratory Study of the Use of Senses, Syntax and Cross-Linguistic Information for Subjectivity Detection in Spanish
An Exploratory Study of the Use of Senses, Syntax and Cross-Linguistic Information for Subjectivity Detection in Spanish
- Open Access
SEMA: an Extended Semantic Evaluation Metric for AMR
Abstract Meaning Representation (AMR) is a recently designed semantic representation language intended to capture the meaning of a sentence, which may be represented as a single-rooted directed acyclic graph with labeled nodes and edges. The automatic evaluation of this structure plays an important role in the development of better systems, as well as for semantic annotation. Despite there is one available metric, smatch, it has some drawbacks. For instance, smatch creates a self-relation on the root of the graph, has weights for different error types, and does not take into account the dependence of the elements in the AMR structure. With these drawbacks, smatch masks several problems of the AMR parsers and distorts the evaluation of the AMRs. In view of this, in this paper, we introduce an extended metric to evaluate AMR parsers, which deals with the drawbacks of the smatch metric. Finally, we compare both metrics, using four well-known AMR parsers, and we argue that our metric is more refined, robust, fairer, and faster than smatch.
Open AccessSEMA: an Extended Semantic Evaluation Metric for AMR
Abstract Meaning Representation (AMR) is a recently designed semantic representation language intended to capture the meaning of a sentence, which may be represented as a single-rooted directed acyclic graph with labeled nodes and edges. The automatic evaluation of this structure plays an important role in the development of better systems, as well as for semantic annotation. Despite there is one available metric, smatch, it has some drawbacks. For instance, smatch creates a self-relation on the root of the graph, has weights for different error types, and does not take into account the dependence of the elements in the AMR structure. With these drawbacks, smatch masks several problems of the AMR parsers and distorts the evaluation of the AMRs. In view of this, in this paper, we introduce an extended metric to evaluate AMR parsers, which deals with the drawbacks of the smatch metric. Finally, we compare both metrics, using four well-known AMR parsers, and we argue that our metric is more refined, robust, fairer, and faster than smatch.