ybracke/dtak-transnormer-full-v1

Dataset Card for DTAK-transnormer-full (v1.0)

Dataset Details

Dataset Description

DTAK-transnormer-full is a modified subset of the DTA-Kernkorpus (Deutsches Textarchiv, German Text Archive Core Corpus). It is a parallel corpus of German texts from the period between 1600 to 1899, that aligns sentences in historical spelling with their normalizations. A normalization is a modified version of the original text that is adapted to modern spelling conventions. This corpus can be used to train and evaluate models for normalizing historical German text.

We also publish a more light-weight variant of this dataset that does not contain all annotation layers, but is otherwise identical: DTAK-transnormer-basic.

Refer to DTAK-transnormer-basic for all information about the dataset's content and creation. The only difference between the datasets is that this dataset contains additional properties based on the original text and the normalized text, see Dataset Structure.

DTAK-transnormer-full has been created and can be further modified with the project transnormer-data. The dataset was created in the context of Text+, for the development of the tool Transnormer.

Usage

Note that, unlike DTAK-transnormer-basic, this dataset cannot be directly loaded with datasets.load_dataset. That is also why the Dataset Viewer does not work properly. Refer to transnormer-data on how to load the data or use DTAK-transnormer-basic if you do not need the additional layers.

Dataset Structure

Data Instances

This is an example instance from the dataset:

{
    "basename":"bodmer_sammlung01_1741",
    "par_idx":984,
    "date":1741,
    "orig":"Das Sinnreiche muß darnach reich an Gedancken ſeyn:",
    "norm":"Das Sinnreiche muss danach reich an Gedanken sein:",
    "orig_tok":["Das","Sinnreiche","muß","darnach","reich","an","Gedancken","ſeyn",":"],
    "norm_tok":["Das","Sinnreiche","muss","danach","reich","an","Gedanken","sein",":"],
    "orig_spans":[[0,3],[4,14],[15,18],[19,26],[27,32],[33,35],[36,45],[46,50],[50,51]],
    "norm_spans":[[0,3],[4,14],[15,19],[20,26],[27,32],[33,35],[36,44],[45,49],[49,50]],
    "alignment":[[0,0],[1,1],[2,2],[3,3],[4,4],[5,5],[6,6],[7,7],[8,8]],
    "lang_fastText":"de",
    "lang_py3langid":"de",
    "lang_cld3":"de",
    "lang_de":1.0,
    "norm_lmscore":6.441500186920166
}

Data Fields

See the description on the Dataset Card of DTAK-transnormer-basic for the properties that occur in both variants. The following properties are only included in DTAK-transnormer-full:

• orig_tok: List[str], tokenized version of orig
• norm_tok: List[str], tokenized version of norm
• orig_spans: List[List[int]], index spans of the elements of orig_tok in orig, inner lists have length 2
• norm_spans: List[List[int]], index spans of the elements of norm_tok in norm, inner lists have length 2
• alignment: List[List[int|None]], alignment of the elements of orig_tok and norm_tok, inner lists have length 2

More information

See DTAK-transnormer-basic.

Dataset Card Author

Yannic Bracke, Berlin-Brandenburg Academy of Sciences and Humanities

License

This corpus is licensed under the CC BY-SA 4.0 license.

Citation Information

@misc{dtak_transnormer_full,
    author    = {Yannic Bracke},
    title     = {DTAK-transnormer-full},
    year      = {2025}
    version   = {1.0}
    url       = {https://huggingface.co/datasets/ybracke/dtak-transnormer-full-v1}
}