Datasets:

Modalities:
Image
Text
Formats:
parquet
Languages:
Danish
ArXiv:
DOI:
Libraries:
Datasets
Dask
License:

Add historical danish handwriting dataset

#102

Checklist

  • I have run the test suite using make test and all tests pass
  • I have added/changed a dataset:
    • I have updated descriptive statistics using make update-descriptive-statistics
    • I have bumped the version use make bump-version
  • If I have added a create.py script I have added the script dependencies required to run that script.
  • I have updated the CHANGELOG.md if appropriate
Danish Foundation Models org

Hej @lrahbek !

(going with English, but feel free to switch to Danish)

Great to see the PR! Generally, things look really good. I have a few pointer/questions:

  • I have added a checklist to your comment, it is originally found here, will you check that it is filled out correctly?
  • The data seems to be transcribed, but is it also human-validated / corrected? If so I would love to have it in the readme.
  • You can delete data/historical-danish-handwriting/create_md.py. You should be ablate run it using python src/dynaword/update_descriptive_statistics.py --dataset {name} --force instead
  • Can you add a note on PII information in the data to the datasheet. I assume the content is considered non-sensitive?

Hej!
Great!

  • I have run the update-descriptive-statistics and will update the checklist accordingly
  • I have also added a line in the readme about the transcription (the pages have all been transcribed by humans)
  • I have removed the create_md.py file

In relation to the PII information; the dataset has not been anonymised, and the dataset card on the original dataset (https://huggingface.co/datasets/aarhus-city-archives/historical-danish-handwriting) discloses that it might include sensitive information - however, because the newest data is from 1939 I think it might be okay, but I am waiting on confirmation regarding that.
I'll get back to you when I get the confirmation that it should be okay regarding the PII, and then i'll add the information to the dataset card.

I have added the PII to the readme, it is just the same information as in the original dataset - is it okay or does anything need clarification? :)

Danish Foundation Models org

Perfect! That was essentially what I was hoping to see

KennethEnevoldsen changed pull request status to open
KennethEnevoldsen changed pull request status to merged
Danish Foundation Models org

Hi @lrahbek , again thanks for the PR and wonderful to see Aarhus putting in the time to share their high quality data.

I made a SoMe post on it here, but wasn't able to @ you in it. Feel free to let me know if I should update it:
https://www.linkedin.com/feed/update/urn:li:activity:7374374925203234816/

Sign up or log in to comment