Volumes < 1

[mromanello]

Hi guys 👋

First off, big kudos to you both for this neat suite of tools to work with HTR-united's data.

I'm preparing a dataset that contains page region annotations (Zones) but not OCR groundtruth.

The dataset passes the HTRVX and HTR_United_Metadata_Generator validation without issues, but it fails with HTRUC because volume < 1 in characters and lines (because of missing OCR).

My assumption was that it would be possible to add to HTR-united's catalog a dataset containing OLR but not OCR GT data... but perhaps this is not true? 🤔

Any help is appreciated 🙏 cheers!

[PonteIneptique]

This is a required feature, this should be possible to disable update where numbers are 0, to avoid a HTRUC failure.

As for the dataset, @alix-tz I think we should allow that, but maybe add a new feature such as dataset type ? (not for right now)

[mromanello]

thanks @PonteIneptique for the quick reply 🙏 So for now we live with it, alles klar! I'll make a PR in any case to add our dataset to the HTR-united catalog asap.

Out of curiosity: is this a validity constraint that can (should?) be relaxed in the JSON schema?

[PonteIneptique]

That's a good question. I strongly believe we should not allow 0-valued quantities, but I can see an argument for it. It'll depend on @alix-tz feedback.
I personally feel like it's an HTRUC issue more than anything else (ie do not feed 0 values into the quantity key)

[alix-tz]

Ok, I definitely agree that we should allow HTRUC to pass in such cases, alles klar for me too!

[PonteIneptique]

I hotfixed the schema accordingly, can you check if it works @mromanello

[mromanello]

Thanks @PonteIneptique I've just re-run the GH actions, and the red flag is gone! I'll make soon a PR for adding the dataset to the HTR-united catalog. (This issue can be closed for me).

[PonteIneptique]

I released it as a hotfix, as it should not break anything.
We'll have to see if the HTR-United website does not ignore your data, because I remember at some point putting in a verification for 0 sized datasets...