Closed rhdunn closed 10 months ago
Why do you think VERB is incorrect? (The line between participles and adjectives is very hard to draw in English.)
From https://en.wikipedia.org/wiki/Adjective#Order:
- Qualifier/purpose – final limiter, which sometimes forms part of the (compound) noun (e.g., rocking chair, hunting cabin, passenger car, book cover)
Note that "United" in noun phrases like "United States", "United Kindom", and "United Air" are classified in this treebank as ADJ, not VERB. "Looking" is also ADJ in this treebank.
For the ... noun
, the part of speech in the ...
is an adjective not a verb, as it isn't defining an action ("the stone is rolling down the hill"), but is defining the name of a magazine -- the "Rolling Stone".
And if it is a particle, shouldn't it be PART+NNP
, not VERB+NNP
?
And if it is a particle, shouldn't it be
PART+NNP
, notVERB+NNP
?
Participle, not particle.
I don't think the semantics offers a good test. In UPOS (unlike PTB/XPOS) we don't distinguish names from non-names except nouns within names (PROPN) vs. other nouns (NOUN).
The best test I can think of is that "very" can modify some adjectives, but not verbs. You can say people are very united on something. I don't think you can say "the news is very breaking".
Note that "United" in noun phrases like "United States", "United Kindom", and "United Air" are classified in this treebank as ADJ, not VERB. "Looking" is also ADJ in this treebank.
I suspect Looking/ADJ is an error.
I see. So per https://github.com/globalwordnet/english-wordnet/issues/953 you have (for compound nouns annotated with NNP
):
VERB
, whileADJ
?Where in this case "looking" is an error as you state.
I don't think the type of participle determines VERB vs. ADJ in general (outside of names there are plenty of -ing ADJs) but perhaps the present participles are less likely to become adjectives than past participles.
Note there are other issues involving participles: #102, #355
Hi @rhdunn, I needed help understanding the link to https://github.com/globalwordnet/english-wordnet/issues/953. There, we are discussing the ADJ organization in the English Wordnet, but we still eventually touch on the deeper problem of what should count as ADJ or not in this branch of the Princeton Wordnet. Indeed, we are all aware of the initial motivations, but some can be revised.
@arademaker The link was regarding the "relates an adjective to a verb" and "relates an adjective to a noun" parts of the different proposed relations. I was speculating on what the EWT tagging rules were based on these adjective classifications.
That is specifically regarding the "present" adjectives (based on the present participle form of a verb), where it looks like EWT classifies that as the underlying VERB (with NNP for the PTB XPOS per the noun phrase rules for compound nouns).
Likewise, for "resultant" adjectives (based on the past participle form of a verb), it looks like EWT is classifying these as ADJ (with NNP for the PTB XPOS per the noun phrase rules for compound nouns).
The part in a noun phrase should be an adjective (ADJ+NNP), not a verb (VERB+NNP):