Puntuguese - A Corpus of Puns in Portuguese with Micro-editions

Puntuguese is a corpus of Portuguese punning texts, including Brazilian and European Portuguese jokes. The data has been manually gathered and curated according to our guidelines. It also contains some layers of annotation:

• Every pun is classified as homophonic, homographic, both, or none according to their specific punning signs;
• The punning and alternative signs were made explicit for every joke;
• We also mark potentially problematic puns from an ethical perspective, so it is easier to filter them out if needed.

Additionally, every joke in the corpus has a non-humorous counterpart, obtained via micro-editing, to enable Machine Learning systems to be trained.

Dataset Sources

The puns were collected from three sources: the "Maiores e melhores" web blog, the "O Sagrado Caderno das Piadas Secas" Instagram page, and from the "UTC - Ultimate Trocadilho Challenge" by Castro Brothers on Youtube.

• Repository: https://github.com/Superar/Puntuguese
• Paper: To be announced

Dataset Structure

The dataset provided via Hugging Face Hub contains two tasks: humor recognition and pun location. The first task uses the text and label columns. For pun location, the columns to be used are tokens and labels. An instance example can be seen below:

{
 "id": "1.1.H",
 "text": "Deve ser difícil ser professor de natação. Você ensina, ensina, e o aluno nada.",
 "label": 1,
 "tokens": ["Deve", "ser", "difícil", "ser", "professor", "de", "natação", ".", "Você", "ensina", ",", "ensina", ",", "e", "o", "aluno", "nada", "."],
 "labels": [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0]
}

Dataset Creation

Data Collection and Processing

The data was manually gathered and curated to ensure that all jokes followed our chosen definition of pun by Miller et al. (2017):

"A pun is a form of wordplay in which one sign (e.g., a word or phrase) suggests two or more meanings by exploiting polysemy, homonymy, or phonological similarity to another sign, for an intended humorous or rhetorical effect."

Every selected pun must satisfy this definition. Gatherers were also provided some hints for this process:

• A sign can be a single word (or token), a phrase (a sequence of tokens), or a part of a word (a subtoken);
• The humorous effect must rely on the ambiguity of said sign;
• The ambiguity must originate from the word's form (written or spoken);
• Every pun must have a "pun word" (the ambiguous sign that is in the text) and an "alternative word" (the sign's ambiguous interpretation) identified. If it is not possible to identify both, the text is not considered a pun and should not be included.

Who are the source data producers?

The original data was produced by professional comedians from the mentioned sources.

Bias, Risks, and Limitations

As in every real-life scenario, the data can contain problematic and insensitive jokes about delicate subjects. In this sense, we provide in out GitHub repository a list of jokes that the gatherers, personally, thought to be problematic.

Citation

BibTeX:

@inproceedings{inacio-etal-2024-puntuguese-corpus,
    title = "Puntuguese: A Corpus of Puns in {P}ortuguese with Micro-edits",
    author = "Inacio, Marcio Lima  and
      Wick-Pedro, Gabriela  and
      Ramisch, Renata  and
      Esp{\'\i}rito Santo, Lu{\'\i}s  and
      Chacon, Xiomara S. Q.  and
      Santos, Roney  and
      Sousa, Rog{\'e}rio  and
      Anchi{\^e}ta, Rafael  and
      Goncalo Oliveira, Hugo",
    editor = "Calzolari, Nicoletta  and
      Kan, Min-Yen  and
      Hoste, Veronique  and
      Lenci, Alessandro  and
      Sakti, Sakriani  and
      Xue, Nianwen",
    booktitle = "Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024)",
    month = may,
    year = "2024",
    address = "Torino, Italia",
    publisher = "ELRA and ICCL",
    url = "https://aclanthology.org/2024.lrec-main.1167",
    pages = "13332--13343",
    abstract = "Humor is an intricate part of verbal communication and dealing with this kind of phenomenon is essential to building systems that can process language at large with all of its complexities. In this paper, we introduce Puntuguese, a new corpus of punning humor in Portuguese, motivated by previous works showing that currently available corpora for this language are still unfit for Machine Learning due to data leakage. Puntuguese comprises 4,903 manually-gathered punning one-liners in Brazilian and European Portuguese. To create negative examples that differ exclusively in terms of funniness, we carried out a micro-editing process, in which all jokes were edited by fluent Portuguese speakers to make the texts unfunny. Finally, we did some experiments on Humor Recognition, showing that Puntuguese is considerably more difficult than the previous corpus, achieving an F1-Score of 68.9{\%}. With this new dataset, we hope to enable research not only in NLP but also in other fields that are interested in studying humor; thus, the data is publicly available.",
}

ACL:

Marcio Lima Inacio, Gabriela Wick-Pedro, Renata Ramisch, Luís Espírito Santo, Xiomara S. Q. Chacon, Roney Santos, Rogério Sousa, Rafael Anchiêta, and Hugo Goncalo Oliveira. 2024. Puntuguese: A Corpus of Puns in Portuguese with Micro-edits. In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), pages 13332–13343, Torino, Italia. ELRA and ICCL.