Closed AngledLuffa closed 2 months ago
@nschneid @amir-zeldes I take it, based on the discussion in https://github.com/UniversalDependencies/UD_English-EWT/issues/526 that words such as unprecedented
, unbelievable
, etc are no longer Polarity=Neg
? I tend to agree with @amir-zeldes 's comment that the feature isn't very useful at that point, but I can implement it here if that's the case
Correct, but @amir-zeldes implemented Negated=Yes
in MISC to cover more cases
I'm a little unmotivated to get such a change right when the models we build won't ever try to predict it.
I guess it's a bit of a rehash of a conversation that already happened, but why not make some sort of morphological feature indicating that kind of negation?
The Core Group concluded that the universal feature should be restricted to the most grammatical negation constructions, as opposed to lexical morphemes containing negative meaning (even if they are semiproductive affixes). IOW un- and in- and de- and anti- and so forth can help derive negative words, but they are not the core grammatical system of negation in English, which is pretty much limited to not and no (+ a few pro-forms).
What was the reasoning, though?
Hard to draw a line if you want to include all words with negative meanings. What about lack, hardly, absence, disable, etc.? The feeling was that the FEATS column is mainly for inflectional morphology and function words.
@nschneid updated the PR to only affect not
. There were no no
or yes
words with INTJ
Change based on similar changes discussed elsewhere in the English UD ecosystem, such as https://github.com/UniversalDependencies/UD_English-EWT/issues/526