Abstract Meaning Representation for Brazilian Portuguese

This repository contains all AMR-annotated corpora developed by the Interinstitutional Center for Computational Linguistics (NILC).

For more information about what is AMR and its specific notations, we indicate the AMR guidelines GitHub repository.

Organization

This repository is organized in subdirectories, which contain each individual corpus. All corpora are distributed under the CC-BY-NC-SA license.

OpiSums-PT-AMR

This corpus contains opinion texts from the OpiSums-PT corpus manually annotated in AMR.

AMRNews

This corpus contains news texts from the Folha de São Paulo newspaper manually annotated in AMR.

AMR-LittlePrince

This corpus contains sentences from the Little Prince tale annotated in AMR through alignment from the English version and later manually revised.

---

For more detailed information about each corpus, please read the README file in the specific corpus directory.

Corpus notation

The corpora follow a standard notation to ease the reading of files. A corpus file contains multiple sentences, each with some metainformation, which starts with a hashtag followed by double colons (# ::) and a keyword (id, snt, alignment...). Then, the AMR graph representation in the PENMAN notation is written. An example is shown below:

# ::id Fala-Serio-Mae.Documento_32.1
# ::snt Amei esse livro .
(a / amar-01
      :ARG0 (e2 / eu)
      :ARG1 (l / livro
            :mod (e3 / esse)))

A blank line separates each sentence.

Statistics

Statistics of each corpus can be obtained by running the script stats_amr.py in this way:

python stats_amr.py <corpus_path> #For example: AMRNews/unsplit/amr.txt

Publications

Both OpiSums-PT-AMR and AMRNews are presented and compared in more detail in the following paper, which has been accepted for publication in DELTA and is currently available as in a pre-print format.

@techreport{InacioEtAl2022,
  type = {Preprint},
  title = {The {{AMR-PT}} Corpus and the Semantic Annotation of Challenging Sentences from Journalistic and Opinion Texts},
  author = {In{\'a}cio, Marcio Lima and Cabezudo, Marco Antonio Sobrevilla and Ramisch, Renata and Di Felippo, Ariani and Pardo, Thiago Alexandre Salgueiro},
  year = {2022},
  month = aug,
  doi = {10.1590/1678-460x202255159},
  url = {https://preprints.scielo.org/index.php/scielo/preprint/view/4652/version/4928},
  urldate = {2022-08-31},
  copyright = {All rights reserved}
}

The AMR-LittlePrince corpus is described in:

@inproceedings{anchieta-pardo-2018-towards,
    title = "Towards {AMR}-{BR}: A {S}em{B}ank for {B}razilian {P}ortuguese Language",
    author = "Anchi{\^e}ta, Rafael  and
      Pardo, Thiago",
    booktitle = "Proceedings of the Eleventh International Conference on Language Resources and Evaluation ({LREC} 2018)",
    month = may,
    year = "2018",
    address = "Miyazaki, Japan",
    publisher = "European Language Resources Association (ELRA)",
    url = "https://www.aclweb.org/anthology/L18-1157",
}