{"id":41905,"date":"2026-07-27T16:27:16","date_gmt":"2026-07-27T14:27:16","guid":{"rendered":"https:\/\/www.graviton.at\/letterswaplibrary\/self-promotion-read-the-places-2105-geocoded-real-world-places-from-392-novels-with-per-place-certainty-ratings-and-source-passages-cc-by-sa-4-0\/"},"modified":"2026-07-27T16:27:16","modified_gmt":"2026-07-27T14:27:16","slug":"self-promotion-read-the-places-2105-geocoded-real-world-places-from-392-novels-with-per-place-certainty-ratings-and-source-passages-cc-by-sa-4-0","status":"publish","type":"post","link":"https:\/\/www.graviton.at\/letterswaplibrary\/self-promotion-read-the-places-2105-geocoded-real-world-places-from-392-novels-with-per-place-certainty-ratings-and-source-passages-cc-by-sa-4-0\/","title":{"rendered":"[self-promotion] Read The Places: 2,105 Geocoded Real-world Places From 392 Novels, With Per-place Certainty Ratings And Source Passages (CC BY-SA 4.0)"},"content":{"rendered":"

<\/p>\n

\n

Disclosure: this is my own project \u2014 I built and maintain it.<\/strong><\/p>\n

Source (the data itself):<\/strong> https:\/\/github.com\/markselby9\/readtheplaces.com<\/a> \u2014 one directory per book under \/books<\/code>, each containing book.json<\/code>, waypoints.json<\/code> and source.txt<\/code>.<\/p>\n

Browsable version:<\/strong> https:\/\/readtheplaces.com<\/a><\/p>\n

Scale:<\/strong> 392 novels, 2,105 places, 298 cities.<\/p>\n

Per-place schema<\/strong> (waypoints.json<\/code>), one real record, abridged:<\/p>\n

{ \"id\": \"westminster-doorstep\", \"name\": \"Clarissa's house, Westminster\", \"progressLabel\": \"10:00\", \"character\": \"clarissa\", \"coords\": [-0.1275, 51.4993], \"placeCertainty\": \"inferred\", \"certaintyNote\": \"Woolf never gives an address. The Dalloways live in Westminster within earshot of Big Ben; scholars place the house around Dean's Yard. Sited here as a considered guess, not a fact.\", \"quoteAnchor\": \"Mrs. Dalloway said she would buy the flowers herself.\", \"passage\": \"...\", \"sources\": [...] } <\/code><\/pre>\n
The field worth arguing about is placeCertainty<\/code>. Geocoding fiction is mostly a disambiguation problem: many places are described rather than named (the abbey in The Name of the Rose<\/em> is a northern Italian abbey Eco never names), and the named ones collide constantly. So each record carries what the resolution was based on, and inferred sitings say so in plain English instead of sitting on the map looking like facts. Filter to placeCertainty != \"inferred\"<\/code> and you get a much smaller, much harder subset.<\/p>\n
Waypoints are ordered by narrative progression rather than geography, so it’s usable for route\/sequence work as well as point work.<\/p>\n
How it was built, honestly:<\/strong> candidate mentions are extracted from the text by an LLM pass, then resolved against gazetteer data and checked by hand. Recall on minor mentions is therefore better than precision, and coverage skews heavily to 19th\u201320th century English-language fiction. Treat it as a curated dataset with a machine-assisted first pass, not a gold standard. It is not synthetic \u2014 every record points at a real passage in a real book.<\/p>\n
Licence is CC BY-SA 4.0. Corrections are PRs against the JSON files, or there’s an issue template if you’d rather just report one.<\/p>\n<\/div>\n
   submitted by    \/u\/markselby9 <\/a> 
 [link]<\/a><\/span>   [comments]<\/a><\/span><\/p>