Correct semantics for proper noun constructions in the ERG?

Onto my next dataset :slight_smile: and there are a LOT of proper nouns.

I have questions about what the appropriate constructions are for a number of examples.

  • Firstname Lastname (e.g. “Liz Conrad”) – is this a compound construction with the last name as the head noun?
    • What about if there are three names? How does it get broken up? Does it depend on if the person has “two last names” vs. a middle name and a last name?
    • Either way will “work” I’m sure but I’m curious what the “correct” analysis would be
  • Seattle, Washington – I thought this could be a compound too, but I have the feeling that “Seattle” should be the head noun and in a regular compound I think the second one is always the head(?)
  • German Language – compound, or adjective+noun?
  • Seattle Airport – would the “airport” in this compound be just the normal noun “airport” or is it treated differently when it’s in a proper noun?
  • The National Hockey League – similar to above, this phrase can be built out of “regular” words, but combined they form a proper noun, is there a special treatment of this as far as semantics go? Or would it just be a compound of ‘hockey+league’ with the adjective ‘national’ in front?

I’m curious about the last cases a bit — for @ecconrad ‘s purposes, it seems like saying that League and Airport are proper nouns is the easier path to broad coverage (likewise for German in German language) since named_rel is one place where we can handle unknown entries in generation.

So what is the potential advantage of using the common noun league or airport (or the adjective German) in such cases? Maybe eventually better realization ranking support, with a WordNet-connected ERG?