Anything can be anything in English; see here.
In computational linguistics, they use more finely grained part-of-speech tags than the seven parts of speech schoolchildren are often taught. For example, the NUPOS tagset uses the tag
n-vvn to indicate “past participle as noun”. It gives an example of “the departed”. So all your examples would be of the
n-vvn part-speech tag under their analysis.
The way to read these compound tags of two parts is that the first part before the dash is what it is used “as”, and the second is what it “is”, morphologically speaking.
n is noun,
vvn is past participle, so
n-vvn is a morphological past participle taking on the rôle of a noun in this particular instance.
Note that the incidence of this part of speech per million words of their training corpus is only 50 per million, which is somewhat low compared with many other tags. Some of that is because of how they subdivide things out. For example, “the late lamented’s house” would be identified not as an
n-vvn, but rather as an
ng1-vvn: a past participle being used as a singular possessive noun (subdividing
n for noun,
2 for singular vs plural, and
g meaning genitive case; similar elaboration can be made for
Even if you aren’t doing computational linguistics, it can be illuminating to read through the NUPOS tagset, to see how things get classified under it. Search down for “NUPOS for English”, and check out the table below that. I think you will find many intriguing things there, some related to your question.