I got tired of searching "where was this filmed" and landing on listicles, so I built the map instead. Search a title to see where it was shot, or search a place to see what was made there.
2,538 locations, 4,214 films and series. It is all built from Wikidata, one statement per pin, so every card links back to the source and you can check me.
It is honest about what it does not know, which no listicle is. 938 productions are recorded only at country level and the page says so. Gladiator is Malta and Morocco and nothing finer. Casablanca resolves to California. It also never claims which scene was shot where, because filming records do not carry that.
430 video games are on there too, placed by where they are set rather than shot, because a game is filmed nowhere.
:https://thefilmmap.com/
wikidata is much easier to query while querying natural language is hard
Although normally I'd agree, not so much here. Wikipedia headers are uniform most of the time so checking if a film has a filming section is easy. Then they handily wrap locations in a tag for their own processing purposes.
For example: [[Havana]]
That's about as easy to parse as you can get.
they don’t just wrap locations in brackets, they wrap every single link in brackets so the software knows to make it a link. wikipedia’s big shtick is making articles have tons of link to each other so that is not easy to figure out which one is the location
not to mention examples like Terminator 1:
when they started filming, the set was moved to Los Angeles. following a link in that section to Rambo 2:
clearly, the movie wasn't shot in "Location shooting", and we can't just discard Thailand when placing all of the films dots on a map. similarly, Get Out begins with "[[Principal photography]] began on February 16, 2016."
Yeah? That's all exceedingly easy to the point of boring to handle. Grab everything between double brackets, test against the openstreetmaps data they're using. If it isn't a location, ignore. Done. As for your Thailand example, discarding regions that are too broad is exactly what they're doing already based on a couple movies I looked at. I'd probably still save it for later as it's a known trusted location I can compare against if I ever find another source of data.
What about the Terminator example? That is a location.
i’m not saying they should discard locations that are too broad (which in fact shouldn’t be linked); i’m saying when there’s multiple locations, all should be included.