nschneid
commented
9 years ago Regarding brave.v vs. brave.j/bravery.n: it seems to me that there is an essential difference in the meaning of the ARG1 that is masked by the label "challenge". For brave.v, it has to be a force or situation that the ARG0 person has no control over, but decides to endure, whereas for brave.j, it has to be an act by the ARG0 that demonstrates his/her bravery. This is why you have to add another verb ("sharing") to the Angelina Jolie example. And it seems to me that the two sentences are not quite paraphrases—"Angelina Jolie braved sharing..." does not necessarily characterize her as brave, it merely says that she is voluntarily doing something likely to be unpleasant.
(Regarding "She was brave in the face of bankruptcy": I think "in the face of" could be used for all kinds of adjectives to denote a cause or stimulus—e.g., "She was sad in the face of bankruptcy"—so this is not really evidence about a core argument of brave.j.)
@timjogorman, what do you think?
timjogorman
cbonial
MarthaSPalmer
uhermjakob
Hi all! Here is the list of the rolesets that we still want to unify out of the ongoing de-unification discussion, with reasons why we want to keep these. We have also started on the task of finding more de-unifications for the remainder of the alphabet, and will post those when we have them
Summary of the de-unification:
51 pairs are being deunified (a few were already split), and we are expanding that list (The current list is at the bottom of this post ). We are also splitting another 7 verb senses to get the granularity required for unification: care, question, value, antagonize, compete, favor, know
12 of the 70 deunifications asked for, however, are sets that we think really should be unified. More detailed notes follow the brief list, including Ulf's complaints:
quick list of the proposed unifications
(more detailed notes at bottom)
hope-n/hope-v vs hopeful-j
We feel that almost all of the interpretations of "hopeful" involve some implication of hoping, but are talking out adding a second roleset for "hopeful" to accommodate the "hopeful prospect" instances. We simply posit that all the "hopeful" instances that entail hope should be unified with "hope".
(http://verbs.colorado.edu/propbank/framesets-english-aliases/hope.html)
equal-v vs equal-j
CU sees these as close enough to unify. Ulf noted "equal distribution of wealth", but you can say "their distributions of wealth equal each other", which is very similar. (http://verbs.colorado.edu/propbank/framesets-english-aliases/equal.html)
sympathize, sympathy vs sympathetic-j:
There are some instances of "sympathetic" that might mean "likable", but for English it almost always has some implication/entailment of sympathizing, and the sympathizing/liking distinction is often impossible to make in single-sentence context. Since they so close to mutual entailment, we propose unification. (http://verbs.colorado.edu/propbank/framesets-english-aliases/sympathize.html)
suspect-v / suspicion-n vs suspicious-j / suspect-j
There is a lot of variance in suspect, suspicious, suspicious, but we think this variance is as tied up with which arguments get realized as it is to which lexical item was used; we think it's close enough to unify. (Further discussion in details.) (http://verbs.colorado.edu/propbank/framesets-english-aliases/suspect.html)
elude-v vs elusive-j
Ulf notes a difference in completiveness here. We do agree that "elude" can often be more completive, but it's not 100% completive (you can get sentences like "His name eluded me until I saw his wife"), and would propose that it's close enough to unify. (http://verbs.colorado.edu/propbank/framesets-english-aliases/elude.html)
glory-v vs glory-n/glorious-j
We are separating out the verbal sense from the nominal one. "glorious" and nominal "glory" seem to be (roughly) the same concept, however, and we think they should be in the same roleset.
(http://verbs.colorado.edu/propbank/framesets-english-aliases/glory.html)
clingy vs cling
Ulf notes that clingy is more habitual, CU would note that "cling" can often be habitual too ("watch out, that shirt clings"), and "clingy" can be coerced to non-habitual ("stop being clingy."). Since "cling" is usually a state, this is less of a sharp contrast than other habitual/nonhabitual contrasts, and close enough to unify.
(http://verbs.colorado.edu/propbank/framesets-english-aliases/cling.html)
comfort-v vs comfortable ( and presumably comfort-n):
We do agree that there is a range of meaning from physical to abstract/emotional comfort, but posit that the range of meaning across verbal and nominal comfort is very broad ("I comforted the victim" vs "The comforting warmth of the stove") and seems like more of a difference than the differences between the verbal and adjectival senses ("The comforting warmth of the stove" vs "The comfortable warmth of my coat"). We think they are close enough to unify. (http://verbs.colorado.edu/propbank/framesets-english-aliases/comfort.html)
conclude/conclusive
Conclusive tends to be related to or being a conclusion. We are amending the frame so that "conclusive evidence" has a better argument for "evidence". (http://verbs.colorado.edu/propbank/framesets-english-aliases/conclude.html)
awe/awesome
The sense of awesome that means "evocative of awe" is hard to distinguish from the more hyperbolic version of the same sense. We have separate, additional senses of "awesome" as well. (http://verbs.colorado.edu/propbank/framesets-english-aliases/awe.html)
brave/brave/bravery
We just talked about trying to unify this, and do accept that the prototypical brave-j use is different from the prototypical brave-v use, but remember that these are the same lemma, and annotators will necessarily have to distinguish semantically between whatever two roles we split these into. We think getting a pure semantic distinction between these (which could also apply to "bravery") is actually quite hard; challenge set in the details. (http://verbs.colorado.edu/propbank/framesets-english-aliases/brave.html)
attend/attending
There is a "pay attention to" roleset currently; which we can rename (other discussion). We are adding a note on how to handle "attending physical" instances. (http://verbs.colorado.edu/propbank/framesets-english-aliases/attend.html)
More detailed discussion, with Ulf's notes:
comfort vs comfort-n/comfortable:
But comfortable, when applies to people, means almost exactly that: – "The school cook is such a comfortable person to be around, the children are always happy when she is in the room."
“Comfort” doesn’t imply prior distress any more than “please” implies prior displeasure. The description following the second example could, just as easily, apply to: "His voice was firm but comforting. " "There the ceramic stove radiated a comforting warmth."
Remember that our assertion is not that the emotional comfort and physical comfort senses are the same idea, but that they both can be realized in "comfort" and "comfortable". The fact that the noun "comfort" is also clearly straddling this line supports this idea.
There is a separate question of whether one wants to split comfort (in the sense of providing solace) from comfort (in the sense of providing physical niceness).
Framenet: sympathize-v: FN: Emotion_directed, Others_situation_as_stimulus
sympathy-n: FN: Emotion_directed (we don’t have this alias currently)
sympathetic-j: FN: Emotion_directed
For most English speakers I've talked to, I can't find clear-cut examples of (2) that don't also entail sympathy; I think "sympathize" just bleeds through so much. More importantly, (1) and (2) are almost impossible to discern from context; I suspect that most instances that anyone would read as (2) would also get annotated as (1).
Since it's both usually implicative of sympathizing, and because they are too confusable to do a more elegant solution (like splitting the sympathizing vs simpatico senses), it seems like the best option is to unify these. We can add a note in the roleset for sympathize to not use this for clear-cut uses of "likeable/pleasant/agreeable".
awe/awesome.
CU notes: The slangy version is simply a hyperbolic use of the term, and the verb can be used the same way. There is also a gradient of ambiguity there; "It was an awesome sight" could be read as "it was excellent" or "it was evocative of awe". Wordnet defines only one sense, "inspiring awe or admiration or wonder".
conclude/conclusion.conclusive
conclude-v: FN: Coming_to_believe, Process_end, Activity_finish
conclusion-n: FN: Coming_to_believe
conclusive-j: FN:-
CU: We are modifying "conclusion" to have an argument for "evidence" roles, i.e. debate-ending elements that are not the agent. Otherwise we think these are the same general idea;
Wordnet:
conclusive (forming an end or termination; especially putting an end to doubt or question) "conclusive proof"; "the evidence is conclusive"
Meriam Webser:
1 of, relating to, or being a conclusion
2 putting an end to debate or question especially by reason of irrefutability
cling-v vs clingy-j
CU notes: You don’t use this word for something that doesn’t cling at least occasionally, which really isn’t any different from the habitual aspect on the verb; "that's the pair of pants that clings".
glory-v vs glory-n, glorious-j
CU Notes: We agree: we are not merging the rolesets for "to take great pride" and "high renown". The candidates for unification here are noun and adjective, as in "glorious victory" / "glory of victory". Annnotators would chose between "high renown" and "to take great pride or pleasure" senses.
elude/elusive
same Framenet role: elude-v: FN: Evading, Elusive_goal elusive-j: FN: Evading
CU notes: I’m not sure that ‘elusive’ is particularly weaker than ‘elude’ in this regard. I would say that if something is elusive and then caught, the part where it ended up being caught would need to be specified just as much as for the verb. "The answer eluded me until I remembered that hint you gave me" doesn't entail that it escaped.
hope/hope/hopeful
hope-v: FN: Desiring
hope-n: FN: Desiring
hopeful-j: FN: Attitude_description
CU notes: We are talking about whether to split off a 'promising' roleset for things like 'hopeful prospects'. Constructions like "have hope that" also fit into that kind of pattern, so we'd propose that we definitely need to get these into the same frame file. Personally, many of the examples used so far to "hopeful" entail the desiring frame; "a hopeful young singer on Broadway" seems to mean "a young singer on Broadway with hopes of success", and not "a promising singer on Broadway"; "she is hopeful that they will win a majority, but has many doubts" seems totally acceptable.
Equal/equal:
CU notes: We see no difference here. Each instance of wealth distribution is equal to all other instances of wealth distribution. The distribution of weath for the rich and poor don't equal each other.
suspect(v)/suspicion(n) vs suspicious(j)/suspect(j)
CU notes: I don’t see a scenario where these are not mutually entailing. Often, "suspicious" or adjectival "suspect" don't include the suspicion Arg1, but that amounts to a transitivity difference, and we don't separate rolesets because of transitivity differences.
With arg1 as subject, ‘suspicious’ is more likely to occur intransitively, but we accept all transitivity variations within our rolesets, so this shouldn’t be a problem. Also, suspicious = suspect-j.
Sonoma County has reasonable grounds to suspect that such information is untrue. Vs. Sonoma County has reasonable grounds to be suspicious that such information is untrue.
The man in the alley is suspicious. Vs. The man in the alley is suspected [of being up to something]
Your behavior is suspicious of criminal activity. Vs. Your behavior is suspected to be criminal activity.
brave-v vs brave-j
CU Note: We think it's fine, however, if you say "Angelina Jolie braved sharing such a personal decision."
We did talk about how one would unify this, and accept that the prototypical brave-j use is different from the prototypical brave-v use, but making a replicable semantic distinction between "facing challenges" and "being brave" is very hard. Challenge set (if anyone has rolesets that would make it easy for annotators to distinguish these senses, feel free to provide them): "He braved the weather" "He braved gunfire to rescue his comrades" "He was then given a medal for his bravery" "I admire Susan for her bravery" "She was brave in the face of bankruptcy"
Splits
black, whiten, redden, worsen, dampen, thicken, moisten, gladden, flatten, deepen, deafen, sadden, ripen, widen, broaden, harden, soften, critical/criticism/criticize, specific/specification/specify, effective/effect/ to effect, dead / deaden, faulty/fault, gag/gagworthy, fuss/fussed, fit/fit, oily/oil, crafty/craft, excel/excellent, even/even, done/do06, conversant/converse, signify/significant, specific/specify, determined/determine, gray/grey, green/green, fuzzy/fuzzy, free/free, flat/flatten, empty/empty, dull/dull, dry/dry, down/down, dizzy/dizzy, distance/distant, dirty/dirty, cool/cool, content/content, clear/clear, blind/blind, chill/chilly, bare/bare, exceed, conceptualize