OPEN LINK ↗
// 35m agoOPENSOURCE RELEASE
City-Name Anagrams Script Maps 2.32M Pairs.
This standalone Python program searches Japanese, U.S., and worldwide settlement gazetteers for city-name anagrams, producing reproducible CSVs and provenance metadata. It finds 1,038,396 anagram-partner names among 2,768,607 normalized entries, while executable checks catch plausible-looking LLM mistakes.
// ANALYSIS
The clever part is not sorting letters; it is treating an AI-assisted curiosity as a data-quality problem with explicit sources, validation, and reproducibility.
- –Combines official Japan and U.S. rosters with GeoNames, covering cities, towns, villages, and hamlets.
- –Shows why scale matters: 2.32 million distinct spelling pairs expand to 23.10 million geographic pairs when duplicate place names are counted.
- –Strong examples include Buena Park–Pekanbaru, Bengaluru–Lauenburg, and Martres-Tolosane–Sant’Orsola Terme.
- –The 37.5% figure applies to a broad settlement dataset, not recognizable major cities.
- –The project’s best lesson is methodological: LLM output suggests candidates, but letter-count checks and source validation decide what survives.
// TAGS
city-name-anagramsllmresearchdatasetevaluationdata-tools
DISCOVERED
35m ago
2026-11-05
PUBLISHED
58m ago
2026-11-05
RELEVANCE
6/ 10
AUTHOR
gwern