README.md

metadata

language:
  - gl
pretty_name: GalCoLA
license: cc-by-4.0
task_categories:
  - text-classification
task_ids:
  - acceptability-classification
tags:
  - galician
  - grammar
  - acceptability
  - cola
  - syntax
  - evaluation
  - nlp
size_categories:
  - 10K<n<100K
configs:
  - config_name: default
    data_files:
      - split: train
        path: train.tsv
      - split: validation
        path: dev.tsv
      - split: test
        path: test.tsv

GalCoLA

Dataset Summary

GalCoLA is a Galician grammatical acceptability dataset in TSV format. It is designed for sentence-level binary classification, where each sentence is labeled as grammatically correct or grammatically incorrect.

The dataset brings together CoLA-style adaptations of Galician materials from two previous research settings:

• targeted syntactic evaluation datasets from PROPOR 2022
• control dependency datasets from ACL 2023

GalCoLA contains 17,088 sentences in total.

Dataset Structure

The dataset is distributed in TSV format and includes three splits:

• train
• validation
• test

Each row contains the following columns:

• paper_id: identifier of the source publication or experimental setting
• source_type: identifier of the source subset or linguistic phenomenon
• source_id: identifier of the original example
• condition: sentence condition
  • a = grammatically correct
  • b = grammatically incorrect
• sentence: sentence in Galician
• label: binary acceptability label
  • 1 = grammatically correct
  • 0 = grammatically incorrect

Example

paper_id	source_type	source_id	condition	sentence	label
PROPOR2022	PER_1	1	a	Cociñei o peixe para o comeres tu.	1
PROPOR2022	NUM_7545	3773	a	Os nenos que xogaban onte alí coa outra cativa presentan mal o acto.	1
PROPOR2022	NUM_4719	2360	a	As rapazas que xogaban onte alí teñen fame.	1

Dataset Creation

GalCoLA combines materials adapted to a CoLA-style acceptability format from two main sources.

PROPOR 2022 (A Targeted Assessment of the Syntactic Abilities of Transformer Models for Galician-Portuguese)

This part of the dataset comes from targeted syntactic evaluation materials for Galician. It includes controlled examples focused on:

• gender agreement
• number agreement
• person agreement

The original evaluation items were converted into sentence-level acceptability pairs, where one sentence is grammatically correct and the other is incorrect.

Total number of rows from this source: 15,552.

ACL 2023 (Dependency resolution at the syntax-semantics interface: psycholinguistic and computational insights on control dependencies)

This part of the dataset comes from experiments on control dependencies in Galician. It includes acceptability materials based on:

• proper names
• pronouns

These materials were also adapted into a CoLA-style binary classification format.

Total number of rows from this source: 1,536.

Dataset Statistics

• PROPOR 2022 adaptation: 15,552
• ACL 2023 adaptation: 1,536
• Total: 17,088

Labels

Label	Meaning
0	grammatically incorrect
1	grammatically correct

Intended Uses

GalCoLA can be used for:

• grammatical acceptability classification in Galician
• evaluation of syntactic agreement phenomena
• probing morphosyntactic abilities of language models
• low-resource NLP research for Galician

Limitations

• The dataset is focused on targeted grammatical phenomena and does not cover all types of acceptability judgments in Galician.
• Many examples are controlled or semi-synthetic adaptations derived from experimental materials.
• The dataset is mainly intended for evaluation and analysis rather than broad-coverage training.

License

GalCoLA is released under the Creative Commons Attribution 4.0 International (CC BY 4.0) license.

Users are free to share and adapt the material, provided that appropriate credit is given to the original source.

Usage

Example with datasets:

from datasets import load_dataset

ds = load_dataset("proxectonos/galcola")
print(ds["train"][0])
print(ds["validation"][0])
print(ds["test"][0])

Acknowledgements

This dataset was compiled within the Nós Project, funded by the Ministerio para la Transformación Digital y de la Función Pública - Funded by EU – NextGenerationEU within the framework of the project ILENIA with reference 2022/TL22/00215336.

Citation

If you use this dataset, please cite the original sources.

@inproceedings{garcia-crespo2022-targeted,
  title = {A Targeted Assessment of the Syntactic Abilities of Transformer Models for Galician-Portuguese},
  author = {Garcia, Marcos and Crespo-Otero, Alfredo},
  booktitle = {Proceedings of the International Conference on the Computational Processing of Portuguese (PROPOR 2022)},
  year = {2022},
  publisher = {Springer Nature},
  series = {Lecture Notes in Artificial Intelligence}
}

@inproceedings{de-dios-flores-etal-2023-control,
  author = {de-Dios-Flores, Iria and García Amboage, Juan Pablo and Garcia, Marcos},
  title = {Dependency resolution at the syntax-semantics interface: psycholinguistic and computational insights on control dependencies},
  booktitle = {Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics},
  year = {2023}
}