---
dataset_info:
  features:
  - name: case_id
    dtype: string
  - name: tabular_data
    struct:
    - name: Overskrift
      dtype: string
    - name: Afgørelsesstatus
      dtype: string
    - name: Faggruppe
      dtype: string
    - name: Ret
      dtype: string
    - name: Rettens sagsnummer
      dtype: string
    - name: Sagstype
      dtype: string
    - name: Instans
      dtype: string
    - name: Domsdatabasens sagsnummer
      dtype: string
    - name: Sagsemner
      dtype: string
    - name: Særlige retsskridt
      dtype: string
    - name: Sagsdeltagere
      dtype: string
    - name: Dørlukning
      dtype: string
    - name: Løftet ud af småsagsprocessen
      dtype: string
    - name: Anerkendelsespåstand
      dtype: string
    - name: Politiets journalnummer
      dtype: string
    - name: Påstandsbeløb
      dtype: string
    - name: Sagskomplekser
      dtype: string
  - name: text
    dtype: string
  - name: text_anonymized
    dtype: string
  splits:
  - name: train
    num_bytes: 193530504
    num_examples: 3917
  download_size: 96393615
  dataset_size: 193530504
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---

# Dataset Card for "domsdatabasen"

## Dataset Description

- **Point of Contact:** [Oliver Kinch](mailto:oliver.kinch@alexandra.dk)
- **Size of dataset:** 199 MB

### Dataset Summary

[Domsdatabasen](https://domsdatabasen.dk/) is a database where you can find and read selected judgments delivered by the Danish Courts. 

Each judgment/case consists of tabular data and a case-descriptive PDF. This dataset collects all these cases, with each sample describing a specific judgment/case.

The PDFs are anonymized to protect sensitive information. Therefore, each sample includes two text versions: 
- `text_anon` (with anonymization tags: \<anonym\>"Some sensitive text"\</anonym\>).
- `text` (without anonymization tags).

`text_anon` is read with [Easyocr](https://github.com/JaidedAI/EasyOCR).

`text` is read with [Easyocr](https://github.com/JaidedAI/EasyOCR) or [Tika-python](https://github.com/chrismattmann/tika-python) 
depending on the PDF and the anonymization method used.

`text_anon` will be empty if no anonymization is detected in the PDF.

### Languages

The dataset is available in Danish (`da`).

## Dataset Structure

An example from the dataset looks as follows.

```
{
  "case_id": "id of case/judgment",
  "tabular_data": {
    "Overskift": "some title",
    "Sagstype": "type of case",
    ...
}
  "text": "pdf text",
  "text_anon": "anonymized pdf text"
  "text_len": <number of chars in text>,
  "text_anon_len": <number of chars in anonymized text>
}
```


### Data Fields

- `case_id`: a `string` feature.
- `tabular_data`: a `dict` feature.
- `text`: a `string` feature.
- `text_anon`: a `string` feature.
- `text_len`: an `int` feature.
- `text_anon_len`: an `int` feature.


### Dataset Statistics

#### Size of dataset

With the PDF texts being provided in two versions, `text` and `text_anon`, the total size of all PDF texts is approximately ~199//2 MB.

#### Number of samples

- 3919

#### PDF Text Length Distribution

Statistics based on `text`.

- Minimum length: 192
- Maximum length: 2101736

![image/png](https://cdn-uploads.huggingface.co/production/uploads/61e0713ac50610f535ed2c88/YTBH-nSHd2b4z6LIjeMF-.png)

## Potential Dataset Issues

See [open issues](https://github.com/oliverkinch/doms_databasen/issues).


## Dataset Creation

### Curation Rationale

There are not many large-scale law datasets in Danish.

### Source Data

The dataset has been scraped from [Domsdatabasen](https://domsdatabasen.dk/).

## Additional Information

### Dataset Curators

[Oliver Kinch](https://huggingface.co/oliverkinch) from the [The Alexandra
Institute](https://alexandra.dk/)

### Licensing Information

The dataset is licensed under the [CC0
license](https://creativecommons.org/share-your-work/public-domain/cc0/).