eduagarcia/portuguese_benchmark

Portuguese Benchmark

This a collection of datasets in Portuguese initially meant to train and evaluate supervised language models such as BERT, RoBERTa, etc... It contains 10 datasets and 18 Tasks for Classification (CLS), NLI, Semantic Similarity Scoring (STS) and Named-Entity Recognition (NER).

NER	Classification	NLI	STS
LeNER-Br	HateBR_offensive_binary	assin2-rte	assin2-sts
UlyssesNER-Br-PL-coarse	HateBR_offensive_level
UlyssesNER-Br-C-coarse	brazilian_court_decisions_judgment
UlyssesNER-Br-PL-fine	brazilian_court_decisions_unanimity
UlyssesNER-Br-C-fine	multi_eurlex_pt
harem-default	Portuguese_Hate_Speech_binary
harem-selective
mapa_pt_coarse
mapa_pt_fine
rrip

Datasets Descriptions

LeNER-Br (NER) [Link]

LeNER-Br is a Portuguese language dataset for named entity recognition applied to legal documents. LeNER-Br consists entirely of manually annotated legislation and legal cases texts and contains tags for persons, locations, time entities, organizations, legislation and legal cases. To compose the dataset, 66 legal documents from several Brazilian Courts were collected. Courts of superior and state levels were considered, such as Supremo Tribunal Federal, Superior Tribunal de Justiça, Tribunal de Justiça de Minas Gerais and Tribunal de Contas da União. In addition, four legislation documents were collected, such as "Lei Maria da Penha", giving a total of 70 documents.

assin2-rte and assin2-sts (NLI/STS) [Link]

The ASSIN 2 corpus is composed of rather simple sentences. Following the procedures of SemEval 2014 Task 1. The training and validation data are composed, respectively, of 6,500 and 500 sentence pairs in Brazilian Portuguese, annotated for entailment and semantic similarity. Semantic similarity values range from 1 to 5, and text entailment classes are either entailment or none. The test data are composed of approximately 3,000 sentence pairs with the same annotation. All data were manually annotated.

HateBR_offensive_binary and HateBR_offensive_level (CLS) [Link]

HateBR is the first large-scale expert annotated dataset of Brazilian Instagram comments for abusive language detection on the web and social media. The HateBR was collected from Brazilian Instagram comments of politicians and manually annotated by specialists. It is composed of 7,000 documents annotated according to three different layers: a binary classification (offensive versus non-offensive comments), offensiveness-level (highly, moderately, and slightly offensive messages), and nine hate speech groups (xenophobia, racism, homophobia, sexism, religious intolerance, partyism, apology for the dictatorship, antisemitism, and fatphobia). Each comment was annotated by three different annotators and achieved high inter-annotator agreement. Furthermore, baseline experiments were implemented reaching 85% of F1-score outperforming the current literature dataset baselines for the Portuguese language. We hope that the proposed expert annotated dataset may foster research on hate speech detection in the Natural Language Processing area.

UlyssesNER-Br-* (NER) [Link]

UlyssesNER-Br is a corpus of Brazilian Legislative Documents for NER with quality baselines. The presented corpus consists of bills and legislative consultations from Brazilian Chamber of Deputies. UlyssesNER-Br has seven semantic classes or categories. Based on HAREM, we defined five typical categories: person, location, organization, event and date. In addition, we defined two specific semantic classes for the legislative domain: law foundation and law product. The law foundation category makes reference to entities related to laws, resolutions, decrees, as well as to domain-specific entities such as bills, which are law proposals being discussed by the parliament, and legislative consultations, also known as job requests made by the parliamentarians. The law product entity refers to systems, programs, and other products created from legislation.

brazilian_court_decisions_judgment and brazilian_court_decisions_unanimity (CLS) [Link]

The dataset is a collection of 4043 Ementa (summary) court decisions and their metadata from the Tribunal de Justiça de Alagoas (TJAL, the State Supreme Court of Alagoas (Brazil). The court decisions are labeled according to 7 categories and whether the decisions were unanimous on the part of the judges or not. The dataset supports the task of Legal Judgment Prediction.

harem-default and harem-selective (NER) [Link]

The HAREM is a Portuguese language corpus commonly used for Named Entity Recognition tasks. It includes about 93k words, from 129 different texts, from several genres, and language varieties. The split of this dataset version follows the division made by [1], where 7% HAREM documents are the validation set and the miniHAREM corpus (with about 65k words) is the test set. There are two versions of the dataset set, a version that has a total of 10 different named entity classes (Person, Organization, Location, Value, Date, Title, Thing, Event, Abstraction, and Other) and a "selective" version with only 5 classes (Person, Organization, Location, Value, and Date). It's important to note that the original version of the HAREM dataset has 2 levels of NER details, namely "Category" and "Sub-type". The dataset version processed here ONLY USE the "Category" level of the original dataset. [1] Souza, Fábio, Rodrigo Nogueira, and Roberto Lotufo. "BERTimbau: Pretrained BERT Models for Brazilian Portuguese." Brazilian Conference on Intelligent Systems. Springer, Cham, 2020.

multi_eurlex_pt (Multilabel CLS) [Link]

MultiEURLEX comprises 65k EU laws in 23 official EU languages. Each EU law has been annotated with EUROVOC concepts (labels) by the Publication Office of EU. Each EUROVOC label ID is associated with a label descriptor, e.g., [60, agri-foodstuffs], [6006, plant product], [1115, fruit]. The descriptors are also available in the 23 languages. Chalkidis et al. (2019) published a monolingual (English) version of this dataset, called EUR-LEX, comprising 57k EU laws with the originally assigned gold labels.

mapa_pt_coarse and mapa_pt_fine (NER) [Link]

The dataset consists of 12 documents (9 for Spanish due to parsing errors) taken from EUR-Lex, a multilingual corpus of court decisions and legal dispositions in the 24 official languages of the European Union. The documents have been annotated for named entities following the guidelines of the MAPA project which foresees two annotation level, a general and a more fine-grained one. The annotated corpus can be used for named entity recognition/classification.

Portuguese_Hate_Speech_binary (CLS) [Link]

The dataset is composed of 5,668 tweets. For its annotation, we defined two different schemes used by annotators with different levels of expertise. Firstly, non-experts annotated the tweets with binary labels (‘hate’ vs. ‘no-hate’). Secondly, expert annotators classified the tweets following a fine-grained hierarchical multiple label scheme with 81 hate speech categories in total. The inter-annotator agreement varied from category to category, which reflects the insight that some types of hate speech are more subtle than others and that their detection depends on personal perception. This hierarchical annotation scheme is the main contribution of the presented work, as it facilitates the identification of different types of hate speech and their intersections.

rrip (CLS) [Link]

Rhetorical role identification (RRI) is an NLP task that consists of labeling the sentences of a document according to a given set of semantic functions (rhetorical roles). This task is useful to applications like document summarization, semantic search, document analysis, among others.
In this corpus, we propose to segment petitions into eight rhetorical roles by mainly considering the analytic needs of judge's' offices in Brazil. We present a corpus of 70 petitions comprising more than 10 thousand sentences manually labeled with the proposed rhetorical roles. These petitions were taken from civil lawsuits filed in the court of the Brazilian state of Mato Grosso do Sul (TJMS).

Citations

Citation for each one of the Tasks:

# LeNER-Br
@InProceedings{luz_etal_propor2018,
    author = {Pedro H. {Luz de Araujo} and Te'{o}filo E. {de Campos} and
            Renato R. R. {de Oliveira} and Matheus Stauffer and
            Samuel Couto and Paulo Bermejo},
    title = {{LeNER-Br}: a Dataset for Named Entity Recognition in {Brazilian} Legal Text},
    booktitle = {International Conference on the Computational Processing of Portuguese ({PROPOR})},
    publisher = {Springer},
    series = {Lecture Notes on Computer Science ({LNCS})},
    pages = {313--323},
    year = {2018},
    month = {September 24-26},
    address = {Canela, RS, Brazil},	  
    doi = {10.1007/978-3-319-99722-3_32},
    url = {https://teodecampos.github.io/LeNER-Br/},
}

# assin2-rte, assin2-sts
@inproceedings{real2020assin,
    title={The assin 2 shared task: a quick overview},
    author={Real, Livy and Fonseca, Erick and Oliveira, Hugo Goncalo},
    booktitle={International Conference on Computational Processing of the Portuguese Language},
    pages={406--412},
    year={2020},
    organization={Springer}
}

# HateBR_offensive_binary, HateBR_offensive_level
@inproceedings{vargas2022hatebr,
    title={HateBR: A Large Expert Annotated Corpus of Brazilian Instagram Comments for Offensive Language and Hate Speech Detection},
    author={Vargas, Francielle and Carvalho, Isabelle and de G{'o}es, Fabiana Rodrigues and Pardo, Thiago and Benevenuto, Fabr{'\i}cio},
    booktitle={Proceedings of the Thirteenth Language Resources and Evaluation Conference},
    pages={7174--7183},
    year={2022}
}

# UlyssesNER-Br-PL-coarse, UlyssesNER-Br-C-coarse, UlyssesNER-Br-PL-fine, UlyssesNER-Br-C-fine
@InProceedings{10.1007/978-3-030-98305-5_1,
    author="Albuquerque, Hidelberg O.
    and Costa, Rosimeire
    and Silvestre, Gabriel
    and Souza, Ellen
    and da Silva, N{'a}dia F. F.
    and Vit{'o}rio, Douglas
    and Moriyama, Gyovana
    and Martins, Lucas
    and Soezima, Luiza
    and Nunes, Augusto
    and Siqueira, Felipe
    and Tarrega, Jo{\~a}o P.
    and Beinotti, Joao V.
    and Dias, Marcio
    and Silva, Matheus
    and Gardini, Miguel
    and Silva, Vinicius
    and de Carvalho, Andr{'e} C. P. L. F.
    and Oliveira, Adriano L. I.",
    editor="Pinheiro, Vl{'a}dia
    and Gamallo, Pablo
    and Amaro, Raquel
    and Scarton, Carolina
    and Batista, Fernando
    and Silva, Diego
    and Magro, Catarina
    and Pinto, Hugo",
    title="UlyssesNER-Br: A Corpus of Brazilian Legislative Documents for Named Entity Recognition",
    booktitle="Computational Processing of the Portuguese Language",
    year="2022",
    publisher="Springer International Publishing",
    address="Cham",
    pages="3--14",
    isbn="978-3-030-98305-5"
}
@InProceedings{10.1007/978-3-031-16474-3_62,
    author="Costa, Rosimeire
    and Albuquerque, Hidelberg Oliveira
    and Silvestre, Gabriel
    and Silva, N{'a}dia F{'e}lix F.
    and Souza, Ellen
    and Vit{'o}rio, Douglas
    and Nunes, Augusto
    and Siqueira, Felipe
    and Pedro Tarrega, Jo{\~a}o
    and Vitor Beinotti, Jo{\~a}o
    and de Souza Dias, M{'a}rcio
    and Pereira, Fab{'i}ola S. F.
    and Silva, Matheus
    and Gardini, Miguel
    and Silva, Vinicius
    and de Carvalho, Andr{'e} C. P. L. F.
    and Oliveira, Adriano L. I.",
    editor="Marreiros, Goreti
    and Martins, Bruno
    and Paiva, Ana
    and Ribeiro, Bernardete
    and Sardinha, Alberto",
    title="Expanding UlyssesNER-Br Named Entity Recognition Corpus with Informal User-Generated Text",
    booktitle="Progress in Artificial Intelligence",
    year="2022",
    publisher="Springer International Publishing",
    address="Cham",
    pages="767--779",
    isbn="978-3-031-16474-3"
}

# brazilian_court_decisions_judgment, brazilian_court_decisions_unanimity
@article{Lage-Freitas2022,
  author = {Lage-Freitas, Andr{'{e}} and Allende-Cid, H{'{e}}ctor and Santana, Orivaldo and Oliveira-Lage, L{'{i}}via},
  doi = {10.7717/peerj-cs.904},
  issn = {2376-5992},
  journal = {PeerJ. Computer science},
  keywords = {Artificial intelligence,Jurimetrics,Law,Legal,Legal NLP,Legal informatics,Legal outcome forecast,Litigation prediction,Machine learning,NLP,Portuguese,Predictive algorithms,judgement prediction},
  language = {eng},
  month = {mar},
  pages = {e904--e904},
  publisher = {PeerJ Inc.},
  title = {{Predicting Brazilian Court Decisions}},
  url = {https://pubmed.ncbi.nlm.nih.gov/35494851 https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9044329/},
  volume = {8},
  year = {2022}
}

# harem-default, harem-selective
@inproceedings{santos2006harem,
    title={Harem: An advanced ner evaluation contest for portuguese},
    author={Santos, Diana and Seco, Nuno and Cardoso, Nuno and Vilela, Rui},
    booktitle={quot; In Nicoletta Calzolari; Khalid Choukri; Aldo Gangemi; Bente Maegaard; Joseph Mariani; Jan Odjik; Daniel Tapias (ed) Proceedings of the 5 th International Conference on Language Resources and Evaluation (LREC'2006)(Genoa Italy 22-28 May 2006)},
    year={2006}
}

# multi_eurlex_pt
@InProceedings{chalkidis-etal-2021-multieurlex,
  author = {Chalkidis, Ilias  
                and Fergadiotis, Manos
                and Androutsopoulos, Ion},
  title = {MultiEURLEX -- A multi-lingual and multi-label legal document 
               classification dataset for zero-shot cross-lingual transfer},
  booktitle = {Proceedings of the 2021 Conference on Empirical Methods
               in Natural Language Processing},
  year = {2021},
  publisher = {Association for Computational Linguistics},
  location = {Punta Cana, Dominican Republic},
  url = {https://arxiv.org/abs/2109.00904}
}

# mapa_pt_coarse, mapa_pt_fine
@article{DeGibertBonet2022,
    author = {{de Gibert Bonet}, Ona and {Garc{'{i}}a Pablos}, Aitor and Cuadros, Montse and Melero, Maite},
    journal = {Proceedings of the Language Resources and Evaluation Conference},
    number = {June},
    pages = {3751--3760},
    title = {{Spanish Datasets for Sensitive Entity Detection in the Legal Domain}},
    url = {https://aclanthology.org/2022.lrec-1.400},
    year = {2022}
}

# Portuguese_Hate_Speech_binary
@inproceedings{fortuna-etal-2019-hierarchically,
    title = "A Hierarchically-Labeled {P}ortuguese Hate Speech Dataset",
    author = "Fortuna, Paula  and
    Rocha da Silva, Jo{\~a}o  and
    Soler-Company, Juan  and
    Wanner, Leo  and
    Nunes, S{'e}rgio",
    editor = "Roberts, Sarah T.  and
    Tetreault, Joel  and
    Prabhakaran, Vinodkumar  and
    Waseem, Zeerak",
    booktitle = "Proceedings of the Third Workshop on Abusive Language Online",
    month = aug,
    year = "2019",
    address = "Florence, Italy",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/W19-3510",
    doi = "10.18653/v1/W19-3510",
    pages = "94--104"
}

#rrip
@inproceedings{aragy_rhetorical_2021,
    series = {Lecture {Notes} in {Computer} {Science}},
    title = {Rhetorical {Role} {Identification} for {Portuguese} {Legal} {Documents}},
    isbn = {978-3-030-91699-2},
    doi = {10.1007/978-3-030-91699-2_38},
    booktitle = {Intelligent {Systems}},
    publisher = {Springer International Publishing},
    author = {Aragy, Roberto and Fernandes, Eraldo Rezende and Caceres, Edson Norberto},
    editor = {Britto, André and Valdivia Delgado, Karina},
    year = {2021},
    pages = {557--571},
}