I think that would be hard to realise (and much work) because : how the computer know where the ocean ends and where the other starts when their's know border between them?
[spoilerjOuch]If you're dyslexic/similar, then apologies for telling you this that you can't make use of, but the use of that <space>colon<space> could have been intended to be several other variations, "theirs" is the relevant possessive but you meant "there's", the contraction, anyway with the negative "no" following it, and it is "wouldn't", later. I'm guessing you're a native English speaker, but not
trying to use slang English, because the drifting from the usual (mis)use is odd.
Just FYI, and only if you can/want to make use of it. Never mind if it's otherwise.[/spoiler]
Named borders are named by people, based upon local perspectives. A desert boundary or a lake shore is a heavy hint to the naming parties, but where there's a blend between two dissimilar zones, or a contact between "that lump and this lump of similar land" one would have to work out what artificial pinch-point might be added to the 'obvious' boundaries around most of the rest of each zone.
I would be tempted to identify 'merging' masses (such as bays, esturies, etc, as gradated offshoots to a deeper sea; seas as coastal areas between land-masses but connected to a surrounding ocean, etc) and find pointy bits along each "Zone1/2" border with the "Definitely not Zone1/2" neighbours that create a convex bit between two concavities (each definitely Zone 1
or Zone 2, respectively) that can be paired with another such pointy bit on the other side.
That would deal with bays (e.g. the line from the tip of the Llŷn Peninsula to one of the points off of Pembrokeshire handily 'encompasses' Cardigan Bay, Wales, but within Cardigan Bay there's smaller 'scalloped' bays and various other features such as the Mawddach estuary) but the more troublesome would be more external features, at least which (unmatched) headlands form boundaries between angled-away concavities, still connecting with adjacent heads that form further corners, beyond which there's no reasonable argument that the sub-zone can extend into the largest mass.
The worldgen process actually has far more information about zone extents than theoretical observers upon the ground would, like oceanic depths and actual biome boundaries (possibly which areas of virtually indistinguishable dry land or which areas of water derived from which
separate spawning features, before blending) so as to emulate anthropogenic naming needs perhaps some fudging to be not
too accurate.
But I'm not so intimately aware of the generating process that I can be sure it can be
properly implemented, and it would probably need tuning, so that "little gap between two rocks that happened to fall off the same eroding cliff" isn't given a naming significance equal to "the inland sea clearly delineated by causeway-like headlands and the tidal race between" sort of features. In RL terms, that is.