Library: Is there a way to check whether a full item ID is valid / whether something is an item ID as opposed to an author ID?

[nschneid]

https://acl-anthology.readthedocs.io/py-v1.0.0/api/utils.ids/ lists RE_ITEM_ID, but that is just for locating the item within the collection, not the full ID. I am looking for a regex that would match "2024.findings-emnlp.355", "P19-1140", and maybe also volumes like "2024.findings-emnlp", "P19-1".

[pratikrath126]

Hey! For matching the full item ID (like 2024.findings-emnlp.355) vs author IDs, you might need to combine the existing RE_ITEM_ID with the collection and volume prefixes. Since the anthology IDs have transitioned from the old letter-based format (like P19-1140) to the newer year-based one, a more comprehensive regex would be something like ^([A-Z]\d{2}-\d{4}|(\d{4}).[a-z0-9.-]+.\d+)$. This targets the structure of both legacy and modern IDs. Author IDs typically don't have the dot-delimited structure or the specific legacy year codes. Hope this helps refine your search!

[nschneid]

@mbollmann can a regex or validation function go in utils.ids?

[mbollmann]

Alternatively, you could also just call RE_COLLECTION_ID.match() and assume that it's an item ID if it matches.

>>> RE_COLLECTION_ID.match("2022.findings-emnlp.42")
<re.Match object; span=(0, 13), match='2022.findings'>
>>> RE_COLLECTION_ID.match("nathan-schneider")
>>>

[nschneid]

I think it's important to document the form of the item IDs somehow, because there will be use cases e.g. in ingestion or form field validation. Maybe even RE_OLD_STYLE_ITEM_ID vs. RE_NEW_STYLE_ITEM_ID because only the latter will apply to new ingestions. If I were to guess I'd probably get it subtly wrong.

[mbollmann]

I’m not sure I see the use case for ingestion or in any other script that directly uses the library. The design philosophy is that the library validates all inputs automatically when making data changes, and I see manual validation of e.g. IDs as an anti-pattern I’d want to discourage.

Validation on the website is a different matter, sure.

[nschneid]

Would you prefer to include in the regex in the docs, then?

[mbollmann]

Maybe, I’ll think about what the best place for this is. For the record, they would look something like

RE_OLDSTYLE_FULL_ID = r"[A-Z][0-9]{2}-[0-9]{4}"
RE_NEWSTYLE_FULL_ID = r"[0-9]{4}\.[a-z0-9]+-[a-z0-9]+\.[a-z0-9]+"