Cupertino Creep hits DC GOP

« previous post | next post »

When I was interviewed for Spiegel Online earlier this week about the dastardly Cupertino effect, I was asked if I thought spellchecker-enabled miscorrections would eventually vanish as spellchecking technology becomes more accurate in predicting potential errors. I said I thought Cupertinos would continue to be with us in one form or another, in large part because of the proper name problem: a reasonably restrictive spellchecker dictionary can never encompass all the proper names that might appear in a given text, particularly unusual foreign names. Consider the old Obama/Osama tangle: after 9/11, Osama was added to Microsoft's spellchecker dictionary, but at the time no one could have predicted that Obama would also be an important name to include. Thus they had to scramble to add Obama when he rose to prominence and spellcheckers were giving Osama as the first suggestion.

Now, as if on cue, the District of Columbia Republican Committee kindly illustrates my point in a new press release.

Wonkette reports that the DC GOP sent out the following release:

Friday the 13th Brings Horrors to the District

Washington, DC: The DC Republican Committee made the following statement in regard to yesterday’s news that the FBI raided the DC’s Chief Technology Office and arrested one District employee and a well known contractor.

“It’s ironic that as the FBI was raiding Vive Kendra’s former District office, when Kendra was giving a speech at FOSE about changing the way the government procurement should be handled. Kendra made national news last night because of his mismanagement of the District’s Procurement Office. He should step down immediately from his position in the Obama administration,” stated DC Republican Committee Chairman Robert J. Kabel.

Who's this Vive Kendra character? Turns out it's actually Vivek Kundra. Kundra was born in New Delhi and moved to Tanzania (where he learned Swahili) before coming to the U.S. Unless he becomes much more famous, we wouldn't expect either his first or last name to be included in a spellchecker dictionary, as opposed to Vive (the French interjection) or the common girl's name Kendra. Thus Vive and Kendra are the first suggestions given by the Microsoft Word spellchecker for his two names. Perhaps, as Barrack Bema's chief information officer, he can do something about the insidious national disease of Cupertino Creep.

Share:



35 Comments »

  1. Robert Coren said,

    March 13, 2009 @ 1:58 pm

    My curmudgeonly opinion is that anybody who blindly accepts whatever a spellcheck program tells them, without looking to see if it's what they really meant, deserves what they get. (I note with slight amusement that whatever spell-checker is being applied to this entry doesn't like "spellcheck" above.)

  2. CWV said,

    March 13, 2009 @ 2:04 pm

    A college roommate once submitted a final take-home exam, in a class on Buddhism, that — after a hastily run spell-check — referred twenty-something times to Siddhartha Guatemala. Not surprisingly, he did poorly.

  3. Terry Collmann said,

    March 13, 2009 @ 2:52 pm

    "… the common girl's name Kendra …

    Not this side of the Atlantic it isn't.

  4. Dan T. said,

    March 13, 2009 @ 3:31 pm

    I remember in the early 1990s looking at online message boards (on proprietary services like Prodigy) that were full of teenage girls, and common names then and there included Kendra and Megan. I presume that there are a lot of twenty- and thirtysomethings with those names now as the trendy-name bulge advances in age.

  5. Karen said,

    March 13, 2009 @ 3:35 pm

    They Might Be Giants have a little song called "Kendra McCormick" so it must be a (relatively) well-known name!

  6. Benjamin Zimmer said,

    March 13, 2009 @ 3:45 pm

    @Terry Collmann: Not even for those familiar with Dumbledore's mother?

    According to the Social Security Administration, Kendra is #202 on the list of most popular names for girls born in the U.S. in 2007. But that's down from #170 in 1999.

    Ah, here's some more complete information from BabyCenter. Kendra peaked in 1987 when it hit #77.

  7. Enosson said,

    March 13, 2009 @ 3:47 pm

    According to the Social Security Administration (http://www.ssa.gov/cgi-bin/babyname.cgi) Kendra was the 202nd most popular female name in 2008, and had ranged from 170th to 227th in the 8 previous years.

  8. Benjamin Zimmer said,

    March 13, 2009 @ 3:51 pm

    @Enosson: Great minds…

  9. Nathan said,

    March 13, 2009 @ 4:39 pm

    I share Robert Coren's curmudgeonly opinion, but I confess to being baffled as well. Is this really what's happening? Are people just turning loose their spellcheckers on their rough text, leaving the software to "correct" everything without human intervention? How could anyone get in such a habit? Wouldn't virtually every document end up with a few hellacious cupertinos?

  10. dr pepper said,

    March 13, 2009 @ 4:46 pm

    Back in the 60's, i heard a story about a man who gave the draft of his psychology thesis, a case study of a manic depressive, to a typist who thought he meant "maniac".

  11. Benjamin Zimmer said,

    March 13, 2009 @ 4:47 pm

    @Nathan: This is the great mystery of the Cupertino effect, which is baffling even to members of the MS Office Natural Language Team who I've talked to. I think it would require some serious research into behavioral psychology to understand how this happens with such great frequency.

  12. D.O. said,

    March 13, 2009 @ 6:10 pm

    Can there be a partial solution of the problem?
    Spell-checkers can just refuse to correct automatically proper names (capitalized words in the middle of the sentence). Especially, when there is a great chance that it is a personal name (I have no idea how to define this, but it might be possible). They may flag the unknown words, but do not react to "change all" option.

  13. Mark F. said,

    March 13, 2009 @ 6:31 pm

    So, under AutoCorrect, Word has the option "Automatically use suggestions from the spelling checker," which I have always turned off for fear of Cupertinos. But I tried turning it on and typing "Vivek Kundra" and I got the wavy underline but no automatic change. It must have some threshold edit distance that can't be exceeded if it is to do the autocorrect.

    Actually, though, I can see how this happens. A round of spell checking is like one of those psychological experiments where you have to click the left button if you see a blue shape or a triangle, and the right button if you see a red shape or a circle, or what have you. It's easy to click the wrong thing. I can easily imagine clicking "Change all" when I meant to click "Ignore".

    under Autocorrect to automatically accept take the first suggestion, but I tried typing in Vivek Kundra and it wasn't changed automatically. So I'm not sure what that check box is supposed to do, but at any rate I would have to specifically click "Change" to make it change.

  14. Mark F. said,

    March 13, 2009 @ 6:32 pm

    Of course, perhaps the reason I can imagine making editing mistakes is that I'm prone to doing things like accidentally leaving in partial paragraphs that were supposed to be replaced.

  15. Dan T. said,

    March 13, 2009 @ 8:34 pm

    A lot of people seem to become drooling idiots when they're at a computer keyboard… for instance, in my work on the back ends of websites that sometimes take input of names, addresses, and the like, I find a distressing number of people who are incapable of typing their own e-mail address without prepending a "www." before it, as if it were a Web site address (which don't necessarily have "www." in them either, but at least it's common there). I eventually start wishing that such errors would commonly lead to fatal accidents, so that Darwinian natural selection can operate to rid the human gene pool of such idiocy.

  16. Skullturf Q. Beavispants said,

    March 13, 2009 @ 10:04 pm

    As a mathematics graduate student, one of my colleagues had "orientable matroids" changed by a spell-checker into "oriental matrons".

  17. Thierry Fontenelle said,

    March 13, 2009 @ 10:44 pm

    Mark F. is right: the Office spell-checker does not automatically replace Vivek Kundra with anything else. It is hard to understand why some people blindly select the suggestion offered by the speller. If this was a case of automatic correction, I could understand some people would be tempted to blame the spell-checker, but this is not the case at all here. The speller does its job in flagging a word that is not included in its lexicon, in offering the closest alternative in terms of edit distance and leaving it up to the user to right-click on the red-squiggled word and click on the suggestion or ignore it.

    Concerning the checkbox Mark is referring to (“Automatically use suggestions from the spelling checker”), it allows you to benefit from automatic replacements when the speller is highly confident that the top suggestion created by the edit distance algorithm is most certainly what the user wanted to type. So, if you type infromation, for instance, with a permutation of “r” and “o”, the speller will automatically change it to information if that checkbox is ticked, because there is only one very likely alternative. Uncheck the box, type infromation, and you will just see the red squiggly line under the misspelled word. Users can of course decide to not use that feature. In the case of Vivek Kundra, there is no such automatic replacement, so the user must have voluntarily chosen the suggestion offered by the speller.

    Concerning proper names and D.O.’s proposal, it is probably a question of trade-off. A large number of users definitely find it useful when the speller catches typos even in proper names. Preventing them from correcting a typo in a capitalized word might irritate them as well (imagine you type Benjanin instead of Benjamin, or Thiery instead of Thierry, and the speller does not allow you to correct it because it is capitalized; I am sure we would get a lot of complaints, wouldn’t we? ;-)

    Thierry [MSFT]

  18. Kenny V said,

    March 13, 2009 @ 10:50 pm

    Today while I was writing, I noticed that Word didn't think "therefrom" was a word. Guess what the first suggestion was—"wherefrom"! I was dumbfounded.

  19. mollymooly said,

    March 14, 2009 @ 12:23 am

    Eleanor Gould, copy editor at the New Yorker, once changed "raunchy" to "paunchy" in a James Thurber piece.

  20. Dick Margulis said,

    March 14, 2009 @ 6:28 am

    Regarding the putative stupidity of the person who accepted "Vive Kendra," this is presumably the same person who sent out a statement in which the first "sentence" was “It’s ironic that as the FBI was raiding Vive Kendra’s former District office, when Kendra was giving a speech at FOSE about changing the way the government procurement should be handled." I suppose that may be a mistranscription by Wonkette, but it is easier to assume Wonkette just copied and pasted. You'd think that with all the recent layoffs, the DC Republican Committee could find someone competent to take the flack job.

  21. Ellen K. said,

    March 14, 2009 @ 8:25 am

    @Thierry Fontenelle

    I don't see anyone blaming spell checkers. Noting and commenting on their involvement in the process, yes. Blaming, no.

  22. Theo Vosse said,

    March 14, 2009 @ 11:00 am

    About correcting proper names… back in 1991 I had already published a simple solution to this problem and similar forms of hypercorrecting (in my native Dutch, there are some spelling errors that make the wrong word look like a compound. However, it does require that you scan the document for token and lemma frequencies before marking words as errors. You build a simple statistical model of proper name probability, and you don't automatically correct it when it's too likely that it's a proper name. However, Word's spell check architecture is still the same as it was back then, and unless that gets changed, Cupertinoes (isn't that the proper plural?) are here to stay.

  23. Theo Vosse said,

    March 14, 2009 @ 11:01 am

    Forget the part about compounds: it's an editing error…

  24. D.O. said,

    March 14, 2009 @ 3:11 pm

    @Thierry Fontenelle: Being an adjunct instructor in a university and a high school teacher before that, I know that people will complain no matter what. I think that flag vs. "change all" distinction makes sense however. I, for one, check each and every word underlined by MSWord in a writing of any consequence (I also check their style suggestions, which are outright horrible in most of the instances), but this is clearly not for everybody.
    You have a lucky name, Thierry. Mine would be changed to Dirty Dostoevsky.

  25. Skullturf Q. Beavispants said,

    March 14, 2009 @ 3:20 pm

    My condolences, Dmitry (I'm guessing).

  26. Thierry Fontenelle said,

    March 14, 2009 @ 7:46 pm

    @D.O. You are right, D.O., the distinction between “flag” and “change all” makes sense. Note that Word offers it, since, in foreground spelling mode (F7), the user can choose between “Ignore Once” (=ignore the current flag), “Ignore All” (=ignore all the instances of that mistake in the current document), “Change Once” (= replace with the suggestion in this case only) and “Change All” (=replace all occurrences of the misspelled word with the selected suggestion). In background spelling mode (where the red squiggly lines appear under misspelled words as you type), you can select the suggestion, or decide to “Ignore” or “Ignore All”. So there is a lot of flexibility and it is of course the user’s responsibility to exercise his or her judgment before ignoring or accepting the suggestions offered by the speller (since, as we have seen, the issue discussed in this post is not due to an automatic replacement).

    Best,

    Thierry [MSFT]

  27. ambrosen said,

    March 14, 2009 @ 8:24 pm

    @Thierry Fontenelle: regarding D.O.'s problem with proper names, and indeed as a minor point of linguistic interest, I relatively often find myself writing letters concerning people whose names are variant spellings on common names, and then of course, what I want is not only the option for that version to be made correct, but for the traditional spelling of the name to be invalid for that document.

    I guess it is only proper names which are so domain specific that they can be correct in one document and incorrect in another.

  28. Barbara Phillips Long said,

    March 14, 2009 @ 9:12 pm

    @Mark F. —
    I often hit the wrong key on the spell checker after a long run of words that are fine; then I click "ignore" once too often.

    If spell-check software had a button for "review previous" that let people go back one or two steps, that would help a lot with the problem. I keep scratch paper and a pen by the keyboard when I check spelling so I can make notes about things I miss when I click before thinking, but a redo option would be much more efficient.

    [Off topic, but I'd like to vent. I do not understand why spell-check software is so bad at handling homonyms. Why can't they query known homonyms and offer short definitions so people can choose the right word?]

    Regarding the name Kendra, go to http://www.babynamewizard.com to see a graph of when the name has been popular in the U.S. and, under the NameMapper feature, to see where it was popular in the U.S. I think some names have world maps, but it has been a while since I spent much time playing with the various features on the site.

    {Is there any linguistics site providing similar graphics about language? I would think data from DARE might be amenable to such treatment. Other than the DARE map, I didn't see any graphics, just text.}

  29. sleepnothavingness said,

    March 15, 2009 @ 8:25 am

    Those of us with slightly longer memories can recall the scourge of the Copy Typist. We had one who would regularly miscorrect "timelines" to "timeliness", and did the same trick with "deadlines", making for droll proof-reading of costed proposal documents.

  30. Terry Collmann said,

    March 15, 2009 @ 5:40 pm

    Back to Kendra – it's certainly not a UK top 100 given name now, and according to the Yournotme website there are only 555 Kendras in the whole UK, against, say, 237,294 called Karen – so no, it's NOT a common name here …

  31. Matt said,

    March 15, 2009 @ 10:17 pm

    @ Terry Collmann:

    Given this particular problem is clearly based in the US, using a spell-checker with US settings, how is it in the least bit relevant that Kendra is less popular in the UK than the US?

    The point was that it was a common enough name that might therefore appear in a spell-checker. Discussing whether it is popular in the UK as well (not to mention that you simply said 'this side of the Atlantic, as if we are all meant to know which side you refer to as 'this') is entirely pointless.

  32. Doug Sundseth said,

    March 16, 2009 @ 1:52 pm

    Since Terry is not a common name on this side of the Atlantic*, you should have known….

    * I'm fairly sure that statement is true for some value of "this side of the Atlantic", though I'm not sure what sort of polygon you'd have to use.

  33. Ken Grabach said,

    March 18, 2009 @ 3:54 pm

    Thierry Fontenelle's comments are apt, that one can accept a spelling suggestion or reject it, or even choose among a group of suggested options. For personal names this comes up frequently. Not surprisingly, my own surname was not originally recognized by my spell checker. This demonstrated to me another function of spell checkers, or at least Microsoft's iteration: one can add items to the dictionary. I did this so that my own name would not be flagged as a potential spelling error every time I checked something, which would have annoyed me beyond toleration.

  34. Beth said,

    March 20, 2009 @ 11:35 am

    In addition to a "review previous" button, I have long wished we had the option to get all editing/proofreading comments in one list after the whole document has been checked, rather than being forced to review each "error" in the order in which it occurs. Word could number the sentences and group similar "errors" together with ref to the sentence numbers. A compiled error list would be useful for writers, and it would be a great teaching tool.

  35. Merri said,

    March 24, 2009 @ 11:06 am

    Notice that using the presence of a capital letter to diagnose a proper name -hence not subjectto correction- wouldn't work in German.
    Perhaps syntactic criteria would vork better.

RSS feed for comments on this post

Leave a Comment