Hugging Face's logo

Datasets:
taln-ls2n
/
CASIMIR

Modalities:
Tabular
Text
Formats:
json
Languages:
English
Size:
10M - 100M
Libraries:
Datasets
pandas
Polars
License:
Dataset card Data Studio Files
xet
 Community
CASIMIR / README.md
JourdanL's picture
JourdanL
Update README.md
e6134f9 verified
preview code
|
Raw
History Blame Contribute Delete
5.53 kB
metadata 
license: mit
language:
  - en
configs:
  - config_name: article_pairs
    data_files:
      - split: train
        path: article_pairs_train.jsonl
      - split: validation
        path: article_pairs_dev.jsonl
      - split: test
        path: article_pairs_test.jsonl
  - config_name: mapping
    data_files:
      - split: train
        path: mapping_train.jsonl
      - split: validation
        path: mapping_dev.jsonl
      - split: test
        path: mapping_test.jsonl
  - config_name: mapping_full_corpus
    data_files: mapping_full_corpus.jsonl
  - config_name: article_pairs_small_test
    data_files: article_pairs_small_test.jsonl
  - config_name: mapping_small_test
    data_files: mapping_small_test.jsonl
  - config_name: metadata
    data_files: metadata.jsonl
  - config_name: reviews
    data_files: reviews.jsonl

CASIMIR: A Corpus of Scientific Articles enhanced with Multiple Author-Integrated Revisions

About

This repository contains the CASIMIR dataset, a dataset of the multiple revised versions of 15,646 scientific articles from OpenReview, along with their peer reviews. Pairs of consecutive versions of an article are aligned at sentence-level while keeping paragraph location information as metadata for supporting future revision studies at the discourse level. Each pair of revised sentences is enriched with automatically extracted edits and associated revision intention.

The creation process of this dataset is described in the following references:

Content

The dataset is composed of different subsets:

  • article_pairs: Main datas, pairs of articles aligned at sentence level with their edits extracted. Divided in train, validation, test.
  • mapping: Mapping of the OpenReview forum id with the associated articles versions id. Divided in train, validation, test.
  • mapping_full_corpus: Same data as previous point but for the whole corpus at once.
  • article_pairs_small_test and mapping_small_test: As running inference on large models over the full test set is computationally expensive and time-consuming, we also provide a smaller test subset, representing 30% of the original test set (approximately 3% of the full dataset, i.e. 468 papers).
  • reviews: All comments posted on the article's forum on OpenReview.
  • metadata: dates, authors, keywords, venue, ids, ...

The dataset article_pairs subset is divided into the following three splits:

Split # articles # versions/article (avg) # edits/pair (avg) Edit length (avg) % (Label) Content % (Label) Grammar-Typo % (Label) Format % (Label) Language
Train 12 488 3.49 141.85 34.90 41.99 22.69 20.44 14.88
Test 1561 3.51 142.98 35.24 42.70 21.43 20.47 15.40
Validation 1597 3.49 143.38 34.35 41.13 24.35 19.79 14.73

The following data fields are available:

  • WIP

Please cite this work as: 

@inproceedings{jourdan-etal-2024-casimir,
    title = "{CASIMIR}: A Corpus of Scientific Articles Enhanced with Multiple Author-Integrated Revisions",
    author = "Jourdan, L{\'e}ane  and
      Boudin, Florian  and
      Hernandez, Nicolas  and
      Dufour, Richard",
    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.257/",
    pages = "2883--2892",
    abstract = "Writing a scientific article is a challenging task as it is a highly codified and specific genre, consequently proficiency in written communication is essential for effectively conveying research findings and ideas. In this article, we propose an original textual resource on the revision step of the writing process of scientific articles. This new dataset, called CASIMIR, contains the multiple revised versions of 15,646 scientific articles from OpenReview, along with their peer reviews. Pairs of consecutive versions of an article are aligned at sentence-level while keeping paragraph location information as metadata for supporting future revision studies at the discourse level. Each pair of revised sentences is enriched with automatically extracted edits and associated revision intention. To assess the initial quality on the dataset, we conducted a qualitative study of several state-of-the-art text revision approaches and compared various evaluation metrics. Our experiments led us to question the relevance of the current evaluation methods for the text revision task."
}