A possible approach:

• For the list of articles we want to generate summaries for, generate a list of all the categories on those pages. e.g. https://en.wikipedia.org/wiki/Special:ApiSandbox#action=query&format=json&prop=categories&list=&titles=Spain%7CParis&formatversion=2&cllimit=500 - be careful about pagination when running the query on multiple pages
• Go over the list of categories with Olga
• Decide which categories to exclude.
• Update the list of pages to filter out ones which feature that category.

@Jdlrobson my original thought to use Wikidata to find entities based on article title, then check if they're human, alive, etc. However, I think the categories approach scales better because it allows us to filter for more arbitrary categories or even parts of categories, like identify anything with a category that matches politic* medic* list of * etc. Not sure about living children, we might have to do some date math for that, but I think your suggested approach gets us most of the way there.

It's going to be a long regular expression. Personally I think we'd need "expanded categories" for every page.

For example, https://en.wikipedia.org/wiki/Nico_Parker is in https://en.wikipedia.org/wiki/Category:English_child_actresses
That category is a subcategory of https://en.wikipedia.org/wiki/Category:British_child_actresses
which is a subcategory of https://en.wikipedia.org/wiki/Category:British_child_actors
which is a subcategory of https://en.wikipedia.org/wiki/Category:European_child_actors
which is a subcategory of https://en.wikipedia.org/wiki/Category:Child_actors_by_continent
which is a subcategory of https://en.wikipedia.org/wiki/Category:Child_actors

It would follow that you should be able to list all articles which have the category 'Child_actors' or a child of that category.

However hitting the API
https://en.wikipedia.org/wiki/Special:ApiSandbox#action=query&format=json&list=categorymembers&formatversion=2&cmtitle=Category%3AChild_actors&cmnamespace=0 the only members are https://en.wikipedia.org/wiki/Child_actor and https://en.wikipedia.org/wiki/Lists_of_child_actors

Would be great if this API allowed you to include pages that were members of subcategories.

... but instead you need to hit the API for every subcategory:
https://en.wikipedia.org/wiki/Special:ApiSandbox#action=query&format=json&list=categorymembers&formatversion=2&cmtitle=Category%3AChild_actors&cmnamespace=14

Hitting the API for every subcategory of every category of every page seems a bit excessive. So far I've added just all categories for every page (not subcategories).

I've put together the following spreadsheet with top-level categories.

https://docs.google.com/spreadsheets/d/1knzKyZT2vCEjCqDjIHEA5MAgSWPG-4CWaK8W9BDhu_0

I haven't done "regulated topics" yet, for the "children" category I just used the year-of-birth categories e.g. "2006 births" combined with the "Living people" category.

So far I have:

• Living people: 1780
• Living politicians: 148
• Living children (born after 2005): 5

We have filtered the list of articles and excluded all living persons and sensitive topics to the best of our abilities:

https://docs.google.com/spreadsheets/d/1knzKyZT2vCEjCqDjIHEA5MAgSWPG-4CWaK8W9BDhu_0/edit?gid=1019355060#gid=1019355060

This exclusion list has been committed to the summary generation repo:

https://gitlab.wikimedia.org/repos/web/simple-summary-server/-/blob/main/excluded_titles.csv?ref_type=heads

And will be checked when summaries are generated and exported.