custmatch
Copyright 2026 Andrew Goodchild

Licensed under the Apache License, Version 2.0 (see LICENSE).

patches/ditto-mps-and-fixes.patch modifies Ditto (https://github.com/megagonlabs/ditto),
Copyright Megagon Labs, licensed under the Apache License, Version 2.0. The patch contains
lines of Ditto's source as diff context.

No third-party datasets are included in the current tree. (Earlier commits in the history
contained the nickname list below and a Wikidata name list, CC0, since removed.) custmatch
downloads it into ~/.cache/custmatch (or $CUSTMATCH_DATA) when first needed:

- nicknames.csv: the carltonnorthern/nicknames list (https://github.com/carltonnorthern/nicknames,
  commit ed160b3), Copyright its contributors, licensed under the Apache License, Version 2.0.
  Downloaded unmodified on first use and checked against its SHA-256.
- profanity_en.txt: the English list of LDNOOBW, the List of Dirty, Naughty, Obscene, and
  Otherwise Bad Words (https://github.com/LDNOOBW/List-of-Dirty-Naughty-Obscene-and-Otherwise-Bad-Words,
  commit 5faf2ba), Copyright its contributors, licensed under CC BY 4.0. Downloaded unmodified
  on first use and checked against its SHA-256; custmatch uses it without the words that are
  real names (custmatch/suspect.py).


Other datasets used by the research scripts are not included either. Scripts download or generate them (FEBRL and
historical_50k via recordlinkage and splink, the Magellan / DeepMatcher and MatchGPT
benchmarks from their own repositories, synthetic people via pseudopeople, name frequencies
from the US Census and Social Security via the FiveThirtyEight data repository (CC BY 4.0)
for the Australian test file), each under its own licence; see docs/reproducing.md. Files in results/ are metrics and logs computed from
them.
