Skip to content

Should a particle at the end of a hyphenated word join the word after it? (Paul Girard-le Brun gives middle 'Girard-le', family 'Brun') #1160

Open

Description

@derek73

parse("Paul Girard-le Brun") gives middle 'Girard-le', family 'Brun'. The family name is Girard-le Brun: two surnames joined by a hyphen, the second of which starts with the particle le. The particle is attached to the word before it, so P2 (particles join forward) never sees it. The same happens:

  • in capitals (Paul GIRARD-LE BRUN)
  • with a longer run after the particle (Paul Girard-de la Tour gives middle 'Girard-de', family 'la Tour')

Measured 2026-11-07 in France's register of elected officials: 12 family names have this shape (X-LE Y, X-DE Y, X-VON Y), and the default reads all 12 wrong. That's out of 909 hyphenated family names.

Questions:

  • Should a listed particle as the last segment of a hyphenated word join forward, as a standalone particle does? A segment test reads part of a word, and the spelling sweep in nameparser/AGENTS.md warns that this path differs from the vocabulary path.
  • Which particles? le, la and de are also French words, but they don't end a hyphenated name word otherwise.

Activity

  1. self-assigned this
    on Nov 8, 2026
  2. added
    designChanges what the model can express (a shape decision)
    on Nov 8, 2026
  3. derek73 commented on Nov 9, 2026

    @derek73
    OwnerAuthor

    Dutch married names have this shape, and it's common: the spouse's surname, a hyphen, then the person's own surname with its particle (Ans Jansen-van der Berg, Noor Bakker-de Vries).

    Measured 2026-11-08 on master 93de2cb:

    • Tweede Kamer open data: 20 of the 22 members whose surname has a particle after the hyphen read it wrong. For example, Ans Jansen-van der Berg gives middle 'Jansen-van', family 'der Berg', which also splits the particle run.
    • ORCID NL: 130 of 134 such family boxes read wrong.
    • ORCID BE: 9 of 12, with a capitalized particle (Lotte Peeters-De Smet gives middle 'Peeters-De', family 'Smet').

    van, de, den and ter are the Dutch particles that end the first half. van also has a given reading (ambiguous), so the question of which particles may end a hyphenated word covers ambiguous ones too.

  4. added
    particlesParticles and surname prefixes (rules/P)
    tokensTokens, separators and punctuation (rules/T)
    and removed on Nov 10, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

designChanges what the model can express (a shape decision)particlesParticles and surname prefixes (rules/P)tokensTokens, separators and punctuation (rules/T)

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions