README.md · maximedb/sick

metadata

dataset_info:
  features:
    - name: pair_ID
      dtype: int64
    - name: sentence_A
      dtype: string
    - name: sentence_B
      dtype: string
    - name: entailment_label
      dtype: string
    - name: relatedness_score
      dtype: float64
    - name: entailment_AB
      dtype: string
    - name: entailment_BA
      dtype: string
    - name: sentence_A_original
      dtype: string
    - name: sentence_B_original
      dtype: string
    - name: sentence_A_dataset
      dtype: string
    - name: sentence_B_dataset
      dtype: string
    - name: SemEval_set
      dtype: string
    - name: label
      dtype: int64
    - name: label_seq2seq
      dtype: string
  splits:
    - name: train
      num_bytes: 1359887
      num_examples: 4439
    - name: validation
      num_bytes: 153417
      num_examples: 495
    - name: test
      num_bytes: 1496660
      num_examples: 4906
  download_size: 822658
  dataset_size: 3009964
license: mit
task_categories:
  - text-classification
language:
  - nl
pretty_name: SICK-NL
size_categories:
  - 1K<n<10K

Dataset Description

Homepage: https://github.com/gijswijnholds/sick_nl
Repository: https://github.com/gijswijnholds/sick_nl
Paper: https://aclanthology.org/2021.eacl-main.126/
Point of Contact: Gijs Wijnholds

Dataset Summary

An automatically translated, manually corrected translation of the SICK dataset of Marelli et al. 2014, intended to boost research in Dutch NLP.

Languages

The dataset is in Dutch.

Dataset Structure

Data Fields

pair_ID: sentence pair ID
sentence_A: sentence A
sentence_B: sentence B
label: textual entailment gold label: entailment (0), neutral (1) or contradiction (2)
relatedness_score: semantic relatedness gold score (on a 1-5 continuous scale)
entailment_AB: entailment for the A-B order (A_neutral_B, A_entails_B, or A_contradicts_B)
entailment_BA: entailment for the B-A order (B_neutral_A, B_entails_A, or B_contradicts_A)
sentence_A_original: original sentence from which sentence A is derived
sentence_B_original: original sentence from which sentence B is derived
sentence_A_dataset: dataset from which the original sentence A was extracted (FLICKR vs. SEMEVAL)
sentence_B_dataset: dataset from which the original sentence B was extracted (FLICKR vs. SEMEVAL)

Data Splits

Train Trial Test 4439 495 4906

Dataset Creation

The dataset was created by first automatically translating all sentences, then by manually correcting any translation errors. This guarantees naturality of the examples while aligning the relatedness scores and entailment labels. Since the data IDs are preserved the dataset is fully aligned on the sentence level.

Additional Information

Licensing Information

This dataset falls under an MIT License.

Citation Information

@inproceedings{wijnholds-etal-2021-sicknl,
    title = "SICK-NL: A Dataset for Dutch Natural Language Inference",
    author = "Wijnholds, Gijs and Moortgat, Michael",
    booktitle = "Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics",
    month = apr,
    year = "2021",
    address = "Online",
    publisher = "Association for Computational Linguistics",
    url = "https://www.aclweb.org/anthology/2021.eacl-main.126/",
}

Contributions

Thanks to @maximedb for adding this dataset.