Wikidata:Property proposal/wikimedia revision identifier
Jump to navigation
Jump to search
wikimedia revision identifier
[edit]Originally proposed at Wikidata:Property proposal/Property metadata
Not done
| Description | identification of Wikimedia page revision |
|---|---|
| Data type | Number (not available yet) |
| Domain | Wikimedia page |
| Allowed values | revision identifier |
| Example | imported from Wikimedia project (P143) → English Wikipedia (Q328) → 1234 |
| Planned use | To specify the exact revision identifier of a Wikimedia page |
| Formatter URL | https://www.wikidata.org/w/index.php?title=Wikidata:Main_Page&oldid=$1, https://en.wikipedia.org/w/index.php?title=Main_Page&oldid=$1 etc. |
| See also | wiki Page Revision ID by Dbpedia |
- Motivation
Currently bots make use of imported from Wikimedia project (P143) to specify the source of information. For example, when some property values are extracted from a English Wikipedia page, imported from Wikimedia project (P143) is used as a property along with the item English Wikipedia (Q328). When it's manually set, property stated in (P248) is used. Currently we are not tracking the exact revision identifier of the Wikipedia page used as a reference for the (extracted) information. The proposed property can be used as a qualifier to specify the exact identifier. Jsamwrites (talk) 15:33, 27 May 2017 (UTC)
Consider the case when information is extracted from a Wikipedia page and fed to Wikidata. imported from Wikimedia project (P143) can be used to specify the Wiki page. Let's assume there are conflicting information concerning a property value in two wiki revisions, A and B. Wikidata editor/Bot extracted the value from Wiki revision A and updated Wikidata. Meanwhile, wikipedia editors has updated the property value (article) in revision B. A user coming to Wikidata must be able to conclude that the property value may be stale since the information was from revision A and Wikipedia has undergone changes and is currently in revision B. The user can update the Wikidata entry, if needed. Jsamwrites (talk) 17:50, 22 August 2017 (UTC)
- Discussion
- Generally, when we use the datatype external-id, we have a formatter url. Is there a formatter url that gives me the revision from a number like 1234 (in the example)? As I see it currently it might be better to store the link to the specific revision that exists at the time the data get's imported. ChristianKl (talk) 21:08, 27 May 2017 (UTC)
- Added the formatter URL. Jsamwrites (talk) 10:09, 28 May 2017 (UTC)
Comment this is not an external identifier because it's not unique. Each Wikimedia project has its own revision identifiers, i.e. the identifier 1234 appears around 600 times in the Wikimedia universe. --Pasleim (talk) 11:21, 28 May 2017 (UTC)- Comment qualifier on sources are technically not possible. --Pasleim (talk) 11:22, 28 May 2017 (UTC)
- Hi Pasleim and ChristianKl, thanks for the feedback. I changed external-identifier to String and removed qualifier. I agree that such a number (e.g. 1234) may appear multiple times on Wikimedia universe. Hence the formatter URL: https://www.wikidata.org/w/index.php?title=Wikidata:Main_Page&oldid=$1. E.g., https://en.wikipedia.org/w/index.php?title=Main_Page&oldid=777395573 points to the version of Wikipedia page (at 12:14, 28 May 2017 (UTC)). However the URL can be further improved based on the value of language concerned, like for English Wikipedia (Q328), https://en.wikipedia.org/w/index.php?title=Main_Page&oldid=777395573. Jsamwrites (talk) 12:14, 28 May 2017 (UTC)
Oppose Wikipedia is not a source. imported from Wikimedia project (P143) is temporary to trace where it came from until it gets replaced by a real source. Multichill (talk) 18:59, 28 May 2017 (UTC)
- The problem is that imported from Wikimedia project (P143) doesn't really tell people where a fact comes from. It could come from many different places on a specific Wikipedia. It doesn't allow a reader to check the provenance of data. This means there's no good way for a bot to tell that if a fact gets removed from a Wikipedia it also should be deleted from Wikidata. It can also make it hard when we find in Wikidata that a value we imported was wrong to go and fix the error in the Wikipedia. ChristianKl (talk) 20:39, 28 May 2017 (UTC)
- I believe that is Multichill's point. It is a pointer that we utilised at a time of the creation of the data; it is an artefact of the process at that time, not the means we should be using now. We can and should do better. — billinghurst sDrewth 02:44, 29 May 2017 (UTC)
- We still import new data from Wikipedia into Wikidata. Of course, it's better to enter data by hand with proper references but in many cases the work of entering data by hand is a completely different project than a quick import. If a person wants to copy the proper citation from EnWiki, they might not know which EnWiki pages holds the citation information. ChristianKl (talk) 11:37, 30 May 2017 (UTC)
- I agree with ChristianKl. There are a lot of tools that are currently available which make use of imported from Wikimedia project (P143) and sadly they are not tracking the exact version of the Wikipedia page. Thereby, we lose the exact source of information. Even considering Multichill's comment that imported from Wikimedia project (P143) is a short term solution, we are still losing the track of very important piece of information regarding the correct permalink page. However, it's not a vague pointer as pointed by billinghurst. Revision identifiers are commonly used even in source code versioning system to point to a particular commit. Luckily all Wikimedia properties give us the revision identifiers and it is left to us to make use of it while developing (import) tools. Jsamwrites (talk) 18:01, 30 May 2017 (UTC)
- The problem is that imported from Wikimedia project (P143) doesn't really tell people where a fact comes from. It could come from many different places on a specific Wikipedia. It doesn't allow a reader to check the provenance of data. This means there's no good way for a bot to tell that if a fact gets removed from a Wikipedia it also should be deleted from Wikidata. It can also make it hard when we find in Wikidata that a value we imported was wrong to go and fix the error in the Wikipedia. ChristianKl (talk) 20:39, 28 May 2017 (UTC)
- Oppose it is redundant to use that process now that we have set the data. We should reference properly, not do vague pointers. This proposal would seem to entrench a practice that is not best practice. — billinghurst sDrewth 02:44, 29 May 2017 (UTC)
Support Rename it to "version in some wiki-based project". d1g (talk) 15:10, 3 June 2017 (UTC)
- It can be used similar to "pages". Not redundant if one want to specify exact version.
- URL formatter could be used to "compare" what was changed.
- Bots would appreciate this property even more than humans because Wikipedia/Wikidata dumps are periodic.
- There is another property Wikimedia database name (P1800), which uses Wikimedia in its name (English). In this way, we are tracking all Wikimedia revision numbers. Jsamwrites (talk) 17:20, 7 June 2017 (UTC)
- Comment Should we use numeric datatype?d1g (talk) 15:10, 3 June 2017 (UTC)
- Hi d1g, I changed the datatype to Integer.
- I also came across the properties wiki Page Revision ID and Wiki Page ID maintained by Dbpedia that serves similar purpose. Jsamwrites (talk) 08:23, 24 June 2017 (UTC)
- Comment I don't think there is a way around using url datatype.
--- Jura 09:00, 24 June 2017 (UTC) - Oppose Use the reference URL property to specify the exact revision when importing. --Izno (talk) 20:39, 26 July 2017 (UTC)
- I think that's what's currently suggested, but I think it would be preferable if these could be filtered out directly by users who don't want them. So a separate property with URL datatype is probably the way to go.
--- Jura 00:26, 29 July 2017 (UTC)- Users can presently do so either by filtering for the specific items or filtering in reference URL for *wiki*. --Izno (talk) 20:30, 20 August 2017 (UTC)
- I think that's what's currently suggested, but I think it would be preferable if these could be filtered out directly by users who don't want them. So a separate property with URL datatype is probably the way to go.
- Comment Consider the case when information is extracted from a Wikipedia page and fed to Wikidata. imported from Wikimedia project (P143) can be used to specify the Wiki page. Let's assume there are conflicting information concerning a property value in two wiki revisions, A and B. Wikidata editor/Bot extracted the value from Wiki revision A and updated Wikidata. Meanwhile, wikipedia editors has updated the property value (article) in revision B. A user coming to Wikidata must be able to conclude that the property value may be stale since the information was from revision A and Wikipedia has undergone changes and is currently in revision B. The user can update the Wikidata entry, if needed. Jsamwrites (talk) 17:50, 22 August 2017 (UTC)
- @Jsamwrites, Izno, D1gggg, Dbpedia, billinghurst, Multichill:@Pasleim, Jura1: Not done, given that there's too much opposition in the current form. Maybe we can make a new trial to get at the same problem of getting more precise sourcing of Wikipedia pages at another time. ChristianKl (talk) 23:36, 18 October 2017 (UTC)