Proposal: also use have-org-role-91 for capitals (of countries)

[uhermjakob]

The special frame have-org-role-91 has served us very well, dramatically improving a wide range of semantic annotations.

I wonder whether we should extend it to capitals of countries and similar city-related concepts that are highly relational to a country (or subdivision thereof).

Example: French president Hollande flew to Rwanda's capital Kigali.

(f / fly-01
  :ARG0 (p / person :name (n / name :op1 "Hollande")
        :ARG0-of (h / have-org-role-91
              :ARG1 (c / country :name (n2 / name :op1 "France"))
              :ARG2 (p2 / president)))
  :destination (c2 / city :name (n3 / name :op1 "Kigali")
        :ARG0-of (h2 / have-org-role-91
              :ARG1 (c4 / country :name (n4 / name :op1 "Rwanda"))
              :ARG2 (c3 / capital))))

Other examples from the corpus: "capital", "main city", "financial center":

• nw.chtb_0177.8 Paoen , the BBC 's reporter stationed in Iraq 's capital , Baghdad , attested on the 13th that no evidence has shown that the air - raid shelter bombed in Baghdad by American bombers was an Iraqi military bunker.
• events_0037.2 The quake struck off the coast at 7:48 a.m. local time, 75 kilometers (50 miles) west of the region's main city of Manokwari, the U.S. Geological Survey said in an alert.
• PROXY_AFP_ENG_20071001_0260.22 20060711 a series of bombs planted on rush-hour trains killed 186 and injured nearly 800 in India's financial center Mumbai.

By contrast, we would not change "town", "village", "port city" etc. as those are much more inherent properties of cities, and not relational.

[nschneid]

I agree that "capital", "financial center", etc. characterize relations, but they do not feel like organization roles to me. How about have-geo-role-91?

headquarter-01 strikes me as appropriate for some institution-location relationships (see #103), but not for relations between two locations.