Is your feature request related to a problem? Please describe.
Hugging Face allows you to mint DOIs for models and datasets, which is awesome overall for citability. But the metadata sent to DataCite is somewhat partial, being largely just the required fields. Users can't provide the richer metadata that the DataCite metadata schema supports to make their works more in various knowledge graphs. They're also blocked from using Hugging Face for models and datasets where institutional or funder policy mandates providing a more complete description of the work.
Describe the solution you'd like
Support the full (or a substantially fuller) use of the DataCite schema when minting and updating DOIs in some form. This could be UI functionality or simply take the form of accepting metadata from the repo, e.g. by way of a yaml or CITATION.cff file (GitHub and Zenodo already support the latter).
Describe alternatives you've considered
What I do for all projects now is mirror models and datasets to Zenodo, where the full DataCite schema is supported, as this is the only way I can meet funder mandates around acknowledgment via use of persistent identifiers. You can see others embedding acknowledgments in the READMEs for funded datasets (e.g. https://huggingface.co/datasets/imageomics/IDLE-OO-Camera-Traps#acknowledgements), but this seems difficult for the funder to monitor, as it requires full text mining.
Additional context
Related to #1208
Is your feature request related to a problem? Please describe.
Hugging Face allows you to mint DOIs for models and datasets, which is awesome overall for citability. But the metadata sent to DataCite is somewhat partial, being largely just the required fields. Users can't provide the richer metadata that the DataCite metadata schema supports to make their works more in various knowledge graphs. They're also blocked from using Hugging Face for models and datasets where institutional or funder policy mandates providing a more complete description of the work.
Describe the solution you'd like
Support the full (or a substantially fuller) use of the DataCite schema when minting and updating DOIs in some form. This could be UI functionality or simply take the form of accepting metadata from the repo, e.g. by way of a yaml or CITATION.cff file (GitHub and Zenodo already support the latter).
Describe alternatives you've considered
What I do for all projects now is mirror models and datasets to Zenodo, where the full DataCite schema is supported, as this is the only way I can meet funder mandates around acknowledgment via use of persistent identifiers. You can see others embedding acknowledgments in the READMEs for funded datasets (e.g. https://huggingface.co/datasets/imageomics/IDLE-OO-Camera-Traps#acknowledgements), but this seems difficult for the funder to monitor, as it requires full text mining.
Additional context
Related to #1208