README.md

---
dataset_info:
- config_name: da
  features:
  - name: text
    dtype: string
  - name: corruption_type
    dtype: string
  - name: label
    dtype: string
  splits:
  - name: train
    num_bytes: 146694
    num_examples: 1024
  - name: val
    num_bytes: 33804
    num_examples: 256
  - name: test
    num_bytes: 276605
    num_examples: 2048
  - name: full_train
    num_bytes: 738488
    num_examples: 5352
  download_size: 703702
  dataset_size: 1195591
- config_name: de
  features:
  - name: text
    dtype: string
  - name: corruption_type
    dtype: string
  - name: label
    dtype: string
  splits:
  - name: train
    num_bytes: 145852
    num_examples: 1024
  - name: val
    num_bytes: 38035
    num_examples: 256
  - name: test
    num_bytes: 295391
    num_examples: 2048
  - name: full_train
    num_bytes: 3784665
    num_examples: 26098
  download_size: 2609370
  dataset_size: 4263943
- config_name: en
  features:
  - name: text
    dtype: string
  - name: corruption_type
    dtype: string
  - name: label
    dtype: string
  splits:
  - name: train
    num_bytes: 142432
    num_examples: 1024
  - name: val
    num_bytes: 33811
    num_examples: 256
  - name: test
    num_bytes: 276421
    num_examples: 2048
  - name: full_train
    num_bytes: 2065205
    num_examples: 15348
  download_size: 1490532
  dataset_size: 2517869
- config_name: fo
  features:
  - name: text
    dtype: string
  - name: corruption_type
    dtype: string
  - name: label
    dtype: string
  splits:
  - name: train
    num_bytes: 146596
    num_examples: 1024
  - name: val
    num_bytes: 39001
    num_examples: 256
  - name: test
    num_bytes: 156312
    num_examples: 1024
  - name: full_train
    num_bytes: 164465
    num_examples: 1148
  download_size: 240821
  dataset_size: 506374
- config_name: is
  features:
  - name: text
    dtype: string
  - name: corruption_type
    dtype: string
  - name: label
    dtype: string
  splits:
  - name: train
    num_bytes: 171843
    num_examples: 1024
  - name: val
    num_bytes: 42652
    num_examples: 256
  - name: test
    num_bytes: 347828
    num_examples: 2048
  - name: full_train
    num_bytes: 680496
    num_examples: 4008
  download_size: 665385
  dataset_size: 1242819
- config_name: nb
  features:
  - name: text
    dtype: string
  - name: corruption_type
    dtype: string
  - name: label
    dtype: string
  splits:
  - name: train
    num_bytes: 129282
    num_examples: 1024
  - name: val
    num_bytes: 30779
    num_examples: 256
  - name: test
    num_bytes: 252311
    num_examples: 2048
  - name: full_train
    num_bytes: 3223287
    num_examples: 25908
  download_size: 2159785
  dataset_size: 3635659
- config_name: nl
  features:
  - name: text
    dtype: string
  - name: corruption_type
    dtype: string
  - name: label
    dtype: string
  splits:
  - name: train
    num_bytes: 123042
    num_examples: 1024
  - name: val
    num_bytes: 30717
    num_examples: 256
  - name: test
    num_bytes: 249767
    num_examples: 2048
  - name: full_train
    num_bytes: 2230281
    num_examples: 18110
  download_size: 1551953
  dataset_size: 2633807
- config_name: nn
  features:
  - name: text
    dtype: string
  - name: corruption_type
    dtype: string
  - name: label
    dtype: string
  splits:
  - name: train
    num_bytes: 137050
    num_examples: 1024
  - name: val
    num_bytes: 35422
    num_examples: 256
  - name: test
    num_bytes: 274470
    num_examples: 2048
  - name: full_train
    num_bytes: 3048722
    num_examples: 22768
  download_size: 2086411
  dataset_size: 3495664
- config_name: sv
  features:
  - name: text
    dtype: string
  - name: corruption_type
    dtype: string
  - name: label
    dtype: string
  splits:
  - name: train
    num_bytes: 142715
    num_examples: 1024
  - name: val
    num_bytes: 36601
    num_examples: 256
  - name: test
    num_bytes: 278198
    num_examples: 2048
  - name: full_train
    num_bytes: 998332
    num_examples: 7434
  download_size: 810940
  dataset_size: 1455846
configs:
- config_name: da
  data_files:
  - split: train
    path: da/train-*
  - split: val
    path: da/val-*
  - split: test
    path: da/test-*
  - split: full_train
    path: da/full_train-*
- config_name: de
  data_files:
  - split: train
    path: de/train-*
  - split: val
    path: de/val-*
  - split: test
    path: de/test-*
  - split: full_train
    path: de/full_train-*
- config_name: en
  data_files:
  - split: train
    path: en/train-*
  - split: val
    path: en/val-*
  - split: test
    path: en/test-*
  - split: full_train
    path: en/full_train-*
- config_name: fo
  data_files:
  - split: train
    path: fo/train-*
  - split: val
    path: fo/val-*
  - split: test
    path: fo/test-*
  - split: full_train
    path: fo/full_train-*
- config_name: is
  data_files:
  - split: train
    path: is/train-*
  - split: val
    path: is/val-*
  - split: test
    path: is/test-*
  - split: full_train
    path: is/full_train-*
- config_name: nb
  data_files:
  - split: train
    path: nb/train-*
  - split: val
    path: nb/val-*
  - split: test
    path: nb/test-*
  - split: full_train
    path: nb/full_train-*
- config_name: nl
  data_files:
  - split: train
    path: nl/train-*
  - split: val
    path: nl/val-*
  - split: test
    path: nl/test-*
  - split: full_train
    path: nl/full_train-*
- config_name: nn
  data_files:
  - split: train
    path: nn/train-*
  - split: val
    path: nn/val-*
  - split: test
    path: nn/test-*
  - split: full_train
    path: nn/full_train-*
- config_name: sv
  data_files:
  - split: train
    path: sv/train-*
  - split: val
    path: sv/val-*
  - split: test
    path: sv/test-*
  - split: full_train
    path: sv/full_train-*
license: cc-by-sa-4.0
task_categories:
- text-classification
language:
- da
- sv
- nb
- nn
- 'no'
- is
- fo
- en
- de
- nl
pretty_name: ScaLA
size_categories:
- 100K<n<1M
---

# Dataset Card for ScaLA

## Dataset Description

- **Point of Contact:** [Dan Saattrup Nielsen](mailto:dan.nielsen@alexandra.dk)

### Dataset Summary

This dataset consists of documents and whether they are grammatically correct or not. It has been automatically generated using [this script](https://github.com/ScandEval/ScandEval/blob/main/src/scripts/create_scala.py), which corrupts documents from a universal dependencies treebank.

### Supported Tasks and Leaderboards

Evaluation of linguistic acceptability (binary classification on correct/incorrect) is the intended task for this dataset. Leaderboards are live [here](https://scandeval.com/).

### Languages

The dataset is available in Danish (`da`), Swedish (`sv`), Norwegian Bokmål (`nb`), Norwegian Nynorsk (`nn`), Icelandic (`is`), Faroese (`fo`), German (`de`), Dutch (`nl`) and English (`en`).

## Dataset Structure

An example from the dataset looks as follows.

```
{
  "text": "some text",
  "corruption_type": <null, "flip_neighbours" or "delete">,
  "label": <"incorrect" or "correct">
}
```

### Data Fields

- `text`: a `string` feature.
- `corruption_type`: a `string` or `null` feature.
- `label`: a `string` feature.


## Dataset Creation

### Curation Rationale

There are very few linguistic acceptability datasets in the given languages.

### Source Data

The dataset has been collected from the [universal dependencies datasets](https://universaldependencies.org/) for the given languages.

## Additional Information

### Dataset Curators

[Dan Saattrup Nielsen](https://huggingface.co/saattrupdan) from the [The Alexandra
Institute](https://alexandra.dk/)

### Licensing Information

The dataset is licensed under the [CC BY-SA 4.0
license](https://creativecommons.org/licenses/by-sa/4.0/).