Fix opus_gnome dataset card

[gojiteji]

I fixed a issue #4805.

I changed "gnome" to "opus_gnome" in README.md.

Fix #4805

[HuggingFaceDocBuilderDev]

The documentation is not available anymore as the PR was closed or merged.

[albertvillanova]

@gojiteji why have you closed this PR and created an identical one?

• document fix in opus_gnome dataset #4807

[gojiteji]

@albertvillanova
I forgot to follow "How to create a Pull" in CONTRIBUTING.md in this branch.

[albertvillanova]

Both are identical. And you can push additional commits to this branch.

[gojiteji]

I see. Thank you for your comment.

[albertvillanova]

Anyway, @gojiteji thanks for your contribution and this fix.

[albertvillanova]

Once you have modified the opus_gnome dataset card, our Continuous Integration test suite performs some tests on it that make some additional requirements: the errors that appear have nothing to do with your contribution, but with these additional quality requirements.

[gojiteji]

the errors that appear have nothing to do with your contribution, but with these additional quality requirements.

Is there anything I should do?

[albertvillanova]

If you would like to address them as well in this PR, it would be awesome: https://github.com/huggingface/datasets/runs/7741104780?check_suite_focus=true

[albertvillanova]

These are the 2 error messages:

E           ValueError: The following issues have been found in the dataset cards:
E           README Validation:
E           The following issues were found for the README at `/home/runner/work/datasets/datasets/datasets/opus_gnome/README.md`:
E           -	No first-level heading starting with `Dataset Card for` found in README. Skipping further validation for this README.

E           The following issues have been found in the dataset cards:
E           YAML tags:
E           Could not validate the metadata, found the following errors:
E           * field 'language':
E           	['ara', 'cat', 'foo', 'gr', 'nqo', 'tmp'] are not registered tags for 'language', reference at https://github.com/huggingface/datasets/tree/main/src/datasets/utils/resources/languages.json

[albertvillanova]

In principle there are 2 errors:

The first one says, the title of the README does not start with Dataset Card for:

• The README title is: # Dataset Card Creation Guide
• According to the template here, it should be: # Dataset Card for [Dataset Name]

[albertvillanova]

In relation with the languages:

• you should check whether the language codes are properly spelled
• and if so, adding them to our languages.json file, so that they are properly validated

[gojiteji]

Thank you for the detailed information. I'm checking it now.

[albertvillanova]

E           ValueError: The following issues have been found in the dataset cards:
E           README Validation:
E           The following issues were found for the README at `/home/runner/work/datasets/datasets/datasets/opus_gnome/README.md`:
E           -	Expected some content in section `Data Instances` but it is empty.
E           -	Expected some content in section `Data Fields` but it is empty.
E           -	Expected some content in section `Data Splits` but it is empty.

[gojiteji]

I added ara, cat, gr, and nqo to languages.json and removed foo and tmp from README.md.
I also write Data Instances, Data Fields, and Data Splits in README.md.

[albertvillanova]

Thanks for your investigation and fixes to the dataset card structure! I'm just making some suggestions before merging this PR: see below.

[gojiteji]

Should I create PR for config.json to add ara cat gr nqo first?
I think I can pass this failing after that.

Or removing ara, cat, gr, nqo, foo, tmp from README.md.

[albertvillanova]

My suggestions for the language codes.

[albertvillanova]

This is a deprecated language tag: it should be replaced by tyj.

However, tyj should be added to our languages.json:

"tyj": "Tai Do; Tai Yo",

See reference here: https://iso639-3.sil.org/request/2015-019

[albertvillanova]

When there exist several codes (ISO 639 1, 2, 3) for the same language, we just use the lower one. For Arabic, we use ar.

As ar already appears in the README, I would remove ara from the README and would not add it to languages.json

Suggested change

"ara":"Arabic",

[albertvillanova]

The same as above: for "Catalan; Valencian" we use ca instead. I would remove cat from README and from here.

Suggested change

"cat": "Catalan; Valencian",

[albertvillanova]

Than language code for Greek is el. I would remove gr from README and from here.

Suggested change

"GR": "Greek",

[albertvillanova]

Once you address these issues, all the CI tests will pass.

[gojiteji]

Thank you for your request, and sorry for the many CI tests runs.

[albertvillanova]

Once the remaining changes are addressed (see unresolved above), we will be able to merge this:

• Remove "ara" from README
• Remove "cat" from README
• Remove "gr" from README
• Replace "tmp" with "tyj" in README
• Add "tyj" to languages.json:

  "tyj": "Tai Do; Tai Yo",
  

[gojiteji]

I did the five changes.

[albertvillanova]

Thanks, @gojiteji for the original fix and all the subsequent improvements to the dataset documentation card.