Should Fritz Freiherr und von Bar read family_base 'Freiherr und von Bar' rather than 'und'? #939

When: a family part is made of particles joined by a connective, as in Fritz Freiherr und von Bar (freiherr, von and bar are all particle vocabulary; und is a conjunction).

Then: the connective is set aside when the part is asked whether every word is particle vocabulary. The part is then all particles, so its words anchor the base. The connective joins them, so it contributes no initial:

• family_base 'Freiherr und von Bar'
• family_particles ''
• initials 'F. F. v. B.'

Because: a word doing a connective's work is never a part's base (rules/P3, rules/R3). Rules/R2 says a part whose words are not acting as particles anchors the base, since "A particle needs a base to attach to".

Today:

• family 'Freiherr und von Bar'
• family_base 'und'
• family_particles 'Freiherr von Bar'
• initials 'F. u.'

The conjunction alone becomes the base of a family that holds three particles. Without und, Fritz Freiherr von Bar reads base 'Freiherr von Bar' and initials 'F. F. v. B.'.

Except:

• A connective that ends the part joins nothing and is a name word (rules/R3). Juan de y keeps family_base 'y' and initials 'J. y.'. If that connective were set aside, de would become the base, which contradicts that reading.

Measured 2026-10-28, counts and token shapes only, over the committed corpus's distinct entries, each read under its declared order. Three families have a base made only of connectives:

• 1 has this shape: particle, connective, particle, particle.
• 1 ends in the connective (particle, connective).
• 1 is a lone connective.

The counts are the same with middle_as_family.

Open: whether the connective stays in the base view. R3's Jon Dough and keeps its connective in the base ('Dough and'), so this draft does the same.

Tracked by the deviates: markers on the example lines in rules/R2 (family_base) and rules/R3 (initials).