Hugging Face's logo

Datasets:
RA-Data-Science
/
DiEm_HTR

Tasks:
Image-to-Text
Languages:
Danish
Size:
10K<n<100K
Tags:
OCR
HTR
handwriting
historical
License:
Dataset card Files
xet
Community
Dataset Viewer
The dataset viewer is not available for this split.
Cannot load the dataset split (in streaming mode) to extract the first rows.
Error code:   StreamingRowsError
Exception:    CastError
Message:      Couldn't cast
text: string
image: null
doc_id: null
sequence: null
alto: null
page: null
to
{'image': Image(mode=None, decode=True), 'doc_id': Value('int64'), 'sequence': Value('int16'), 'alto': Value('string'), 'page': Value('string')}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/utils.py", line 99, in get_rows_or_raise
                  return get_rows(
                         ^^^^^^^^^
                File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
                  return func(*args, **kwargs)
                         ^^^^^^^^^^^^^^^^^^^^^
                File "/src/services/worker/src/worker/utils.py", line 77, in get_rows
                  rows_plus_one = list(itertools.islice(ds, rows_max_number + 1))
                                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 2431, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 1952, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 1984, in _iter_arrow
                  pa_table = cast_table_to_features(pa_table, self.features)
                             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/table.py", line 2192, in cast_table_to_features
                  raise CastError(
              datasets.table.CastError: Couldn't cast
              text: string
              image: null
              doc_id: null
              sequence: null
              alto: null
              page: null
              to
              {'image': Image(mode=None, decode=True), 'doc_id': Value('int64'), 'sequence': Value('int16'), 'alto': Value('string'), 'page': Value('string')}
              because column names don't match

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

Dataset Card for DiEm HTR

The DiEm HTR dataset is a ground truth dataset for historical danish handwriting in the 17th and 18th century, generated as part of the Digitalisering af Enesteministerialbøger project at the Danish National Archives.

Dataset Details

Dataset Description

The Digitalisering af Enesteministerialbøger project (DiEm) at the Danish National Archives aims to transcribe and make publically available all of the danish parish registers from before the 1813-reform by using the Handwritten Text Recognition (HTR) platform Transkribus. To this end, ground truth training data for the HTR-models have been created, which we now make publically available through Hugging Face.

TheDiEm dataset consists of 975 transcribed images, containing a total of 67410 text lines and 383339 words.

The following parish registers are part of the dataset:

doc_id parish_name period events transcribed notes
8010610941-6469506 Jersie Sogn 1747-1814 Contains: Births and baptisms, Confirmations, Marriages, Burials. Volunteers have manually transcribed the book.
8010629211-6471563 Solrød Sogn 1752-1814 Contains: Births and baptisms, Confirmations, Marriages, Burials. Volunteers have manually transcribed the book.
8010634501-6472194 Ølsemagle Sogn 1732-1766 Contains: Births and baptisms, Marriages, Burials. Volunteers have manually transcribed the book. Only the first 41 pages of the book is made Ground Truth, which are the available pages in the dataset.
0834247701 Ål Sogn 1691-1726 Contains: Births and baptisms, Marriages, Burials. Volunteers have manually transcribed the book. The pictures of the book are not the ones that are available on The Danish National Archives, because the DiEm-project took its own pictures.
8010720391-6482219 Bråby Sogn 1645-1749 Contains: Births and baptisms, Marriages, Burials. Volunteers have manually transcribed the book. Not all pages from the book could be made Ground Truth. Only the Ground Truth pages are available in this dataset.
8010785121-6490574 Vester Ulslev Sogn 1722-1801 Contains: Births and baptisms, Confirmations, Marriages, Burials. Volunteers have manually transcribed the book. Not all pages from the book could be made Ground Truth. Only the Ground Truth pages are available in this dataset.
8026583501-9436330 Branderup Sogn 1667-1814 Contains: Births and baptisms, Confirmations, Marriages, Burials. Volunteers have manually transcribed the book. We only made the pages up to and including 1711 Ground Truth from this book. Only the Ground Truth pages are available in this dataset.
8027899071 Jels Sogn 1648-1721 Contains: Births and baptisms, Confirmations, Marriages, Burials. Volunteers have manually transcribed the book.
8033700071-12391767 Vinding Sogn (Vejle Amt) 1651-1749 Contains: Births and baptisms, Marriages, Burials Volunteers have manually transcribed the book. We only made parts of this book Ground Truth because of its bad condition and layout. Pages from 1657 up to and including 1735 is Ground Truth from this book. Only the Ground Truth pages are available in this dataset.
8010560601-22053014 Værløse og Farum Sogne 1781-1814 Contains: Confirmations. An early version of the DiEm-project’s text recognition model in Transkribus recognised the book.
8034392541-22064545 Tranekær Sogn 1812-1818 Contains: Communion og Confirmations. An early version of the DiEm-project’s text recognition model in Transkribus recognised the book.
8034325781-12671556 Ellested og Gislev Sogne 1675-1687 Contains: Births and baptisms, Marriages, Burials. An early version of the DiEm-project’s text recognition model in Transkribus recognised the book.
8034365901-12691596 Nyborg Sogn 1665-1700 Contains: Burials. An early version of the DiEm-project’s text recognition model in Transkribus recognised the book.
8031903941 Hvidbjerg Vesten Å Sogn 1658-1704 Contains: Births and baptisms. An early version of the DiEm-project’s text recognition model in Transkribus recognised the book.
8010717631 Alsted Sogn (Sorø Amt) 1804-1814 Contains: Births and baptisms, Confirmations, Marriages, Burials. An early version of the DiEm-project’s text recognition model in Transkribus recognised the book.
8034383771 Føvling (Ribe Amt) og Holsted Sogne 1758-1772 Contains: Births and baptisms, Burials. An early version of the DiEm-project’s text recognition model in Transkribus recognised the book.

Uses

Direct Use

The dataset is meant primarily as ground truth for training HTR models for 17th and 18th century danish handwriting. A separate dataset will be made available containing the ground truth pages for our region detection model as these only partly overlap with the ground truth for HTR.

Unpacking the parquet file and putting the images in a root folder and the alto/page xmls in subfolders called 'alto' and 'page' will allow import of the transcriptions into the desktop client of Transkribus, if you want to include the dataset as training data in your Transkribus project. We have created a small tool 'UnpackRAParquet' that can help you unpack the parquet-files in the proper structure, which is included in the tools/ subfolder of this dataset. Windows binary: UnpackRAParquet.exe (SHA256: 30A4E7F67EABD275514721C0FD1AEE99F04C6A376CBEA5F027E1E5BD95458AAF)

Out-of-Scope Use

The dataset is not suited for training textline polygon extraction models, as the polygons have been generated by Transkribus and not manually adjusted. The dataset is not suitable for training models for baseline detection either since some blurred/unreadable words in the beginning and ending of a line has been cut out. Some blurred/unreadable lines has also been left unmarked. Finally we advice using the DiEm Regions dataset if you want to train a model to detect text regions, as the regions within the Diem HTR dataset has not been manually validated. The DiEm Regions dataset should be made available here on Hugging Face in the winter 2025-26.

Dataset Structure

Each data instance represents a single scanned, segmented and transcribed image with handwritten text corresponding to either 1 or 2 physical pages in the parish registers. The dataset contains the following fields:

  • image: a jpeg image containing a scan of the original physical page
  • doc_id: internal id of parent parish register (book) to which the page belongs used at the National Archives of Denmark
  • sequence: an incremental id denoting the order of the page within the parent document
  • alto: an xml-encoded string containing layout and content information of the physical page, stored in ALTO XML, version 4
  • page: an xml-encoded string containing layout and content information of the physical page, stored in PAGE XML, version 2013-07-15

To uniquely identify a single page within the dataset, one can use the doc_id integer in combination with the sequence integer.

Dataset Creation

Curation Rationale

The dataset constitutes the ground truth HTR created through the Transkribus interface as part of the DiEm project, managed by the National Archives of Denmark. The project seeks to correctly read all the danish parish registers before 1813. The books included in the ground truth dataset has been selected to represent the variance in handwriting style through the period 1650-1800 as well as possible. Likewise we have tried to include books containing all the events that might be described in parish registers (baptisms, confirmations, marriages and burials).

Source Data

The source data is the parish registers written by parish clerks and priests in the danish parishes in the period ~1650-1800. They parimarily describe the clerical events carried out for the people of the parish, noting among other things names and dates at baptisms, confirmations, marriages and burials.

Data Collection and Processing

Already existing transcriptions of parish records that were transcribed true to the source and therefore suitable to become Ground Truth were collected and manually inserted in Transkribus by the project workers. Since most of the parish records the project needs to text recognize are from the 18th century, most of our training data (around 70%) are from that period. Volunteers in the DiEm-project have proofread all the parish records to Ground Truth in Transkribus following the project’s proofreader guide, so the proofreading is as similar as possible even though several people contributed to it. The project workers assisted the volunteers in the process.

Who are the source data producers?

The transscriptions used initially as ground truth training data has been created by various genealogists in Denmark, often as part of research into their own ancestry. The transscriptions have since been verified and corrected by project workers.

Annotations [optional]

Annotation process

The DiEm-project’s volunteers have used the Unclear-tag in Transkribus whenever they have come across a word that was unreadable. They also marked underlined words, super- and subscript by using Transkribus’ text editor.

Who are the annotators?

Annotations have been created by participants in the DiEm projects at the National Archives of Denmark.

Personal and Sensitive Information

The dataset contains no personal, private or sensitive information as all information is over 200 years old.

Bias, Risks, and Limitations

Be advised that the language used in the text is somewhat limited as it concerns baptisms, confirmations, marriages and burials, and there therefore is a predominance of dates and names when compared to natural language. Most of the training data consists of baptisms (around 42%) followed by burials (around 24%). Around 10% are marriages and 4% confirmations. The rest are chronological kept parish records, where we assume that the grouping of events are roughly the same. Most of the training data are from the 18th century (around 70%) and the rest are from the 17th century.

Recommendations

Users should be aware that:

  • We do not use diacritical signs in the training data. For example, we transcribe the letter ‘ö’ as ø.
  • We mark unreadable words with ‘##’ in the text, and tag them with the Unclear-tag in Transkribus.
  • We use strikethrough, underline, sub- and superscript in the training data.
  • Only the Ground Truth pages from the parish records are a part of this dataset.
  • We always use the sign ¬ when there are hyphenated words in the text.
  • Nine of the parish records are manually transcribed and seven are recognised using early versions of the DiEm-projects text recognition models and corrected afterwards.

Citation [optional]

[N/A]

Glossary [optional]

{{ glossary | default("[More Information Needed]", true)}}

More Information [optional]

Thank you to the Augustinus Foundation for funding the DiEm project.

Dataset Card Contact

Point of Contact: Markus Schunck

Downloads last month
34