Datasets:

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

Add instruct variant on Danish datasets (~1M words)

#40
by KennethEnevoldsen - opened
Danish Foundation Models org
edited 7 days ago

Early models like FLAN-T5 obtained a large degree of training data from existing structured datasets. These can be used to create reasonable instruct samples.

approx sum: ~1M


An outlined examples of Euphemism-of-Death

Intended format could be:

In the following sentence is the  '{target}' an euphemism for death:

{sentence_before}{sentence}{{sentence_after}

e.g.

In the following sentence is `vandrer heden` an euphemism for death:

hendes sidste levedage ; jeg formoder hun vil takke ham før hans godhed og velsigne ham før hun vandrer heden . Bed ham komme herhen , så hurtigt som muligt .  Det var efter modtagelsen af dette brev , at Peter

We could probably make a variant of this (should probably be in Danish too)

estimate of size: 3 sentence of 12 words *1000 cases = 36 000

paper: https://aclanthology.org/2025.coling-main.90.pdf
github: https://github.com/mime-memo/Euphemism-of-Death

KennethEnevoldsen changed discussion title from Add instruct variant Euphemism of Death to Add instruct variant Euphemism of Death (~36k words)
KennethEnevoldsen changed discussion title from Add instruct variant Euphemism of Death (~36k words) to Add instruct variant on Danish datasets (~36k words)
KennethEnevoldsen changed discussion title from Add instruct variant on Danish datasets (~36k words) to Add instruct variant on Danish datasets (~688k words)
KennethEnevoldsen changed discussion title from Add instruct variant on Danish datasets (~688k words) to Add instruct variant on Danish datasets (~1M words)
Danish Foundation Models org

Good idea! I would leave out Danish citizen tests though, to avoid data leakage with evals.

Danish Foundation Models org
edited 6 days ago

Yeah, I did think about that as well. Nordjylland news is also in the test distribution. We could potentially move some datasets to the "test" split.

edit: Looking a bit more into it I had a few questions on the dataset (which I created a discussion on over at the repo: https://github.com/EuroEval/EuroEval/discussions/1181)

Sign up or log in to comment