TiagoLubianaBot

Incorrect edits

Both of these edits are extremely wrong. They added two items for taxonomic genera (genus) to a data item for a taxonomic class. --EncycloPetey (talk) 04:19, 25 May 2023 (UTC)

Thanks! It seems like UMLS-to-MeSH mappings are quite less then perfect. I have stopped the bot for now, and I'll try some other heuristics to clean it up. TiagoLubiana (talk) 10:44, 25 May 2023 (UTC)

Here is another incorrect match; connecting a taxonomic division to a genus. --EncycloPetey (talk) 16:31, 25 May 2023 (UTC)

And another incorrect match. --EncycloPetey (talk) 16:32, 25 May 2023 (UTC)

It should be possible to check whether the target link is a genus and compare to be sure that the WD data item is also ranked as a genus. A genus should only ever be matched to another genus, not to some other rank. For some top botanical ranks, like division, subclass, and class, the rank might not always match, but for orders, families, and genera, the rank should always match. --EncycloPetey (talk) 16:36, 25 May 2023 (UTC)

The latest round of edits look good to me. --EncycloPetey (talk) 15:21, 19 June 2023 (UTC)

Thanks! I changed the approach. The NCBI matches seem to be on point. There might be some incorrect matches lingering, I'll fix them when they appear. TiagoLubiana (talk) 16:33, 19 June 2023 (UTC)

P2892 on parent items

Hello. Looking at the bot's edits to hydrazine (Q58447) here, it has repeatedly placed the CUI numbers for hydrazine compounds under hydrazine, when WD items for each already exists. I've sense moved these IDs to the appropriate items. Does the bot not try to detect for items before making edits? Or does this relate to the above question? — Huntster (t @ c) 13:47, 25 May 2023 (UTC)

Thanks for the notice. It relates with the above, the bot is using MeSH as a proxy for mappings, and sometimes the UMLS-MeSH mappings on the original source I not precise.

After the comments, I've set it to avoid updating the many-to-one mappings, all new edits should be from 1:1.

I am studying removing the previous many-to-one mappings in batch too; even though they are related matches, it might be better to just not having them on Wikidata altogether.

Thoughts are welcome! TiagoLubiana (talk) 14:19, 25 May 2023 (UTC)

Time

Hello. This bot keeps adding wrong identifiers to time (Q11471) [1][2]. I'd expect that manually undone bot edits would not be re-done.

[1] https://www.wikidata.org/w/index.php?title=Q11471&diff=prev&oldid=1902490618

[2] https://www.wikidata.org/w/index.php?title=Q11471&diff=prev&oldid=1902491422 Toni 001 (talk) 06:12, 26 May 2023 (UTC)

Thanks, Toni. While I agree with you in theory, that is quite difficult to modify in the bot code, I don't know how to do it.

That edit is indeed what the original source claimed. Maybe this is one case where instead of removing it, we should tag it with the "deprecated" rank?

Anyways, the MeSH-UMLS batch has been run, and it should not add that again. TiagoLubiana (talk) 10:22, 26 May 2023 (UTC)