Datasets:

elenanereiss

/

german_legal_entity_recognition

like 1

@lhoestq @albertvillanova

@elenanereiss thanks for flagging! Should we just remove this one in favor of https://huggingface.co/datasets/elenanereiss/german-ler then?

Yes, thanks!

Thanks for reporting, @elenanereiss .

I'm just thinking that maybe another possibility could be to redirect this dataset to https://huggingface.co/datasets/elenanereiss/german-ler because this one has already been downloaded 1.45k times. This way, people that have already downloaded it, will be redirected to your dataset next time they try to load their local dataset. I should first check if this is technically possible, though.

Please, let me know what you think.

that's a good idea. the people who used the dataset then know that the dataset still exists. thank you, @albertvillanova

Hi @elenanereiss ,

After a discussion with the rest of the datasets team, we think that the "automatic" redirect is not a good idea in this case because the dataset here and yours are indeed different (as you have pointed out: the splits here are incorrect, involving legal texts from seven federal courts).

We think in this specific case it might be more appropriate to do what we have already done with "multilingual_librispeech": before deleting the old dataset, we keep it during a reasonable amount of time but informing about the newer dataset with:

• a warning in the dataset card pointing to the new dataset: you can see it here or just visiting the landing page of the datset here
• a warning in the code pointing to the new dataset, that appears each time the user loads the old dataset: you can see it here

This way people still using the old dataset are informed about the new one, so that they can gradually switch.

After a prudential time (1 or 2 months?), we would eventually delete the old dataset.

Please, let us know if you agree with this approach.

Hi @albertvillanova ,
I agree. Thanks!👍

@elenanereiss , this dataset has been deprecated: https://huggingface.co/datasets/german_legal_entity_recognition/discussions/3