6.486 Words that are their own opposites

[The Linguist List]

----------------------------------------------------------------------
LINGUIST List:  Vol-6-486. Sun 02 Apr 1995. ISSN: 1068-4875. Lines: 164
 
Subject: 6.486 Words that are their own opposites
 
Moderators: Anthony Rodrigues Aristar: Texas A&M U. <aristar at tam2000.tamu.edu>
            Helen Dry: Eastern Michigan U. <hdry at emunix.emich.edu>
 
Asst. Editors: Ron Reck <rreck at emunix.emich.edu>
               Ann Dizdar <dizdar at tam2000.tamu.edu>
               Ljuba Veselinova <lveselin at emunix.emich.edu>
               Annemarie Valdez <avaldez at emunix.emich.edu>
 
-------------------------Directory-------------------------------------
 
1)
Date: Tue, 28 Mar 1995 23:37-EST
From: Marion.Kee at A.NL.CS.CMU.EDU
Subject: Words that are their own opposites
 
2)
Date: Wed, 29 Mar 1995 00:45:57 -0500 (EST)
From: Jason S Jensen (jsj at acsu.buffalo.edu)
Subject: RE: 6.465 Words that are their own opposites
 
 
3)
Date: Thu, 30 Mar 1995 11:35 +0100 (MET)
From: WERTH at ALF.LET.UVA.NL
Subject: RE: 6.465 Words that are their own opposites
 
-------------------------Messages--------------------------------------
 
1)
Date: Tue, 28 Mar 1995 23:37-EST
From: Marion.Kee at A.NL.CS.CMU.EDU
Subject: Words that are their own opposites
 
 
 
Lee Hartman offers another pair involving a preposition:
 
)  Marion Kee, who works with prepositions, should enjoy this autoantonym:
)FOR (= 'because of' / 'in spite of'):
)  1)  We hired her for her computer skills.
)  2)  For all her faults, she's still a good computer-guruani.
 
"enjoy"?  I'm not sure, but, yes, I've classified this contrast.  1) uses
"for" to introduce a rationale; 2) uses "for" to introduce a concession.
)From the point of view of automatic text analysis, "for" is a nightmare.
If I could remove one preposition from the English corpus I work with,
"for" would be it.  Probably its single most difficult feature is the
fact that it sees use as both a preposition and a clausal connector.
(Other prepositions, e.g. the time indicators "before" "after" and
"until", also share this duality, but in more restricted environments
than "for".)  Running a close second to that dual-use problem is the
multiplicity of meanings and shades of meaning that can be expressed
with "for".  ("to", "with", "on" and "in" are other major culprits in
this regard.)  Some of those meanings are opposites, near-opposites, or
hard enough for me to pin down that I can't always predict in advance
whether or not an antonymous meaning will arise between two particular
classifications.  I *hope* my classification scheme prevents antonymy
WITHIN a given catgory, but sometimes you only discover such problems
via catastrophic failure in the translated text.
 
To illustrate some of the difficulty involved with "for", I'll take Lee's
examples and revise them slightly:
 
  3)  I love her for her faults.  ("because of")
  4)  For all her computer skills, she's a lousy manager. ("in spite of")
 
Here, it's the "faults" that constitute the 'rationale' use of "for", and
the "computer skills" that constitute the 'concession' use of "for".  So
any attempt to automatically ferret out the meaning that's intended
(setting aside for the moment the approximately 15 other meaning
categories I've identified for "for"--see what I mean?  It's insidious!)
cannot rely solely on the knowledge that the preposition is "for" and
the object of the preposition is "faults" or "computer skills".
 
Automatic analysis of the meaning thus has to rely on other cues, such
as sentence structure.  In general, in automatic bottom-up analysis of
a sentence, the "higher up" you get in building the parse tree, the
more likely it is that you've made a mistake somewhere in your analysis
of the sentence structure.  So, you may have a 99% chance of correctly
identifying a prepositional phrase _per se_, but only a 75% chance of
correctly deducing which other word in the sentence it attaches to
(modifies), and so on for correctly analyzing other parts of the
sentence structure and how they relate to each other.  In trying to
meet this challenge you can encounter a sort of "Catch-22" situation:
knowing the meaning expressed by the preposition could help you choose
the correct parse of the sentence, and knowing the correct parse of the
sentence could help you choose the right meaning for the preposition.
It's not hard to see why the use of multiple strategies is a
popular approach (such as combining a syntactically-based grammar with
a semantically-organized knowledge base and some statistically-based
heuristics.)  One way of implementing such an approach is to
base-generate all the parses permitted by the syntactic grammar and
then systematically eliminate the ones that don't fit with the rest of
the available knowledge; another approach is to interweave all these
knowledge sources so that you don't waste computational resources in
generating a bunch of useless parses and then throwing them out.  (This
last is harder to do; aside from implementation problems, it's highly
destructive--you risk never even generating the right structure.)  The
problems presented by prepositional phrase attachment in particular are
very difficult and by no means "solved".  The excessive ambiguity of
English preposition meanings is certainly part of the problem.
 
Given that ambiguity, I'm sure there are other auto-antonymous English
prepositions out there just waiting to be found.  Also, because the little
devils are kind of promiscuous (consorting with verbs and giving birth
to verb+particle combinations; hanging out in the morphology and
latching onto roots) there are probably more forms such as the
"oversight/overlook" pair and the "turn on" example offered in
earlier postings.
 
Does this kind of large-scale prepositional confusion also happen in other
languages?  The only language other than English that I'm reasonably
fluent in is French, and while it has its standouts ("en" comes to mind
immediately), my experience in translating from English into French
suggests that the problem is not as severe in French.  From studying
Russian (too long ago, I'm afraid) I recall that its preposition use is
more restricted and more precise, but I never had any real claim to
fluency in that language, so my judgement is not reliable.  Would
anyone care to comment on this point?
 
 --Marion
 ----------------------
Marion Kee                                         | All opinions are my own;
Knowledge Engineer, Center for Machine Translation | when CMU wants my opinions
Carnegie Mellon University    Pittsburgh, PA, USA  | it pays for them.
 
--------------------------------------------------------------------------
2)
Date: Wed, 29 Mar 1995 00:45:57 -0500 (EST)
From: Jason S Jensen (jsj at acsu.buffalo.edu)
Subject: RE: 6.465 Words that are their own opposites
 
 
I apologize if this point has already been made, but my favorite example
of autoantonymy has always been "oversight" (the act of overseeing, i.e.,
being responsible) vs. "oversight" (the act of overlooking, i.e., if you
like, being irresponsible).
 
Marion Kee, in a recent posting, seems to be comparing "oversight" to
"oversee"; however, "oversight" itself can have these two quite distinct
meanings ("Congressional Oversight Committees" are always worth a
chuckle), due probably to the various meanings of "over" which Kee
describes.  Again, I apologize if the point was made and I just missed it.
 
--------------------------------------------------------------------------
3)
Date: Thu, 30 Mar 1995 11:35 +0100 (MET)
From: WERTH at ALF.LET.UVA.NL
Subject: RE: 6.465 Words that are their own opposites
 
The posting about 'hasta' in Spanish reminds me somehow of a strange dialectal
use of 'while' in East Riding Yorkshire. I was first alerted to it when an old
local inhabitant advised me: "Don't prune your roses while Easter". But the
summum of this use has to be the road sign which was exhibited on the traffic
bridge across the Humber at Goole (in the days before the modern Humber Bridge
was built where the river is much wider). It read: "Do not cross while lights
stop flashing". How's that for an autoantonym?
 
Regards,
Paul Werth
 
--------------------------------------------------------------------------
LINGUIST List: Vol-6-486.