Wikidata:Requests for permissions/Bot/Pi bot 5
The following discussion is closed. Please do not modify it. Subsequent comments should be made in a new section. A summary of the conclusions reached follows.
Approved --Lymantria (talk) 05:32, 12 June 2018 (UTC)
Pi bot (talk • contribs • new items • new lexemes • SUL • Block log • User rights log • User rights • xtools)
Operator: Mike Peel (talk • contribs • logs)
Task/s: Add human name property values by deduction and/or importing from other Wikimedia projects
Code: On bitbucket; see below.
Function details: I would like to run a number of different scripts to bot-populate family name (P734) and given name (P735) values for human (Q5) entries.
The first of these is to add family name (P734) values based on the label minus the given name (P735), avoiding complicated cases by checking for spaces in the residual text. The code for that is on BitBucket, which relies on a script that pre-generates a list of family name QIDs. I've run a 50-item test run (with a minor hiccup at the start that added duplicates, that's been fixed by double-checking for the current presence of a P735 value). This should affect hundreds of thousands items (I'm not sure how many as the query to select them currently times out, but there are ~230,000 at commons:Uses of Wikidata Infobox with no family name at the moment, and there are ~2.5 million uses of P735 compared to ~0.5 million for P734). It will be an initial big run, followed by regular (potentially daily) runs to catch new cases.
Later scripts will import values from the other Wikimedia projects, such as defaultsort labels from commons and elsewhere. This should then be able to work through some of the more complex cases that are skipped over by the first task. If those can also be covered by this task request, that would be good, otherwise I can put them forward separately when the code is ready.
Thanks. --Mike Peel (talk) 17:13, 10 June 2018 (UTC)
Support Code looks good. I like especially that you use the existing given name (P735) value to check if it is an easy structured name. --Pasleim (talk) 17:45, 10 June 2018 (UTC)
New section
[edit]When adding P735 (which is mentioned, but not detailed above), please bear in mind that not all items have labels that actually start with a given name.
--- Jura 10:21, 12 June 2018 (UTC)
- @Jura1: Definitely. I know that Chinese names are often in reverse order. Mostly I'm planning on using the existing defaultsort information to populate that one, and it will only work for cases where there is an item with instance of (P31) = male given name (Q12308941) or female given name (Q11879590) (probably also with a gender check). If you can think of any examples where there could be issues, I'd appreciate them. Thanks. Mike Peel (talk) 12:47, 12 June 2018 (UTC)
- "Kim" is the classical sample. I tend to exclude anything that has less than 4 letters and select by P27.
--- Jura 12:52, 12 June 2018 (UTC)- Good point. I've added code to avoid short surnames, and I'll also include that in the forename code when I write that. Thanks. Mike Peel (talk) 14:35, 12 June 2018 (UTC)
- For given names, almost any names that aren't in the usual script haven't really been looked into, at least as far as I know (notably Arabic, Indian). Korean may be the exception. Occasionally, people add P735 erroneously to Romans. For family names, you might want to exclude Icelandic people .. Personally, I avoided P734, as it's slightly more complicated ..
--- Jura 07:30, 13 June 2018 (UTC)
- "Kim" is the classical sample. I tend to exclude anything that has less than 4 letters and select by P27.