Professor Finds Fault with MS Grammar Checker
ChuckOp writes "
front-page article in the Seattle Post-Intelligencer states: "The University
of Washington associate professor has embarked on a one-man mission to persuade the Redmond company to improve the grammar-checking function in its popular word-processing program. Sandeep
Krishnamurthy is also trying to raise public awareness of the issue." He includes some twisted prose that the grammar checker fails to find fault with, such as: "Marketing are bad for brand big and small. You Know What I am Saying?" and "Gates do good marketing job in Microsoft". This last comment is disputed by retired Microsoft researcher Karen Jensen, who developed part of the underlying technology; "Only by knowing that 'Gates' probably refers to Bill Gates -- and not to the plural of the movable portion of a fence -- would the program know to suggest using 'does' instead." The professor also has several twisted examples available."
By visiting his site, I found out that he is the Associate Professor of Marketing and E-Commerce, and I was played right into his hand and visited his site! I bet he's laughing with his colleague from the Department of Statistics right now.
Rock that crushes, Paper & Scissors that don't matter.
That's unpossible!
I mean come on. I'm not even a MS fan and I agree that their product could get better, but if you're going to write like a 4 year old... And it would be different if the product's purpose was souly to check grammar. It's NOT. There's a point at which the user has to step in and use some sense and actually EDIT their work themselves.
The eternal struggle of good vs. evil begins within one's self.
Ms Jensen doesn't note that the example is STILL incorrect even if one doesn't assume Gates is a proper noun. Grammatically, it should be, "Gates do good marketing jobs in Microsoft." Plural JOBS.
Of course, the chances of seeing a Jobs in Microsoft these days are probably nil.
These people looked deep into my soul and assigned me a number based on the order in which I joined.
I agree that there are many shortcomings in Microsoft's grammar checker. However, to what extent should we bother trying to improve it? English is an extraordinarily complex language and it should be easy to construct "twisted" examples which any grammar checker would miss; any standard intro AI course will warn of the dangers of overfitting data anyway. On the other end of the spectrum, I'm sure it's easy to construct examples which the grammar checker will never allow but which are often perfectly acceptable under certain circumstances. English grammar simply isn't as black and white as, say, C syntax, no matter what we geeks would like. :-)
I'd rather have a program that points out the typical mistakes that occur when you cut and paste around, i.e. phrases without a verb, or with too many verbs, than one that is giving false alarms all the time. A grammar checker cannot fix a bad writer. Neither a spell checker, for that matter. (Do you write "advise" or "advice"?)
Personally, I don't use grammar checkers (not available for Emacs AFAIK anyway), and a spell checker only if I doubt about a particular word. There are way too many words in the kind of things that I write that make the spell checker freak out.
BTW, I probably made a mistake or two in this posting. My excuse is that I ain't no native speaker. :)
Avantslash: low-bandwidth mobile slashdot.
If the MLA would come up with a formal specification of the English language that was a recursively enumerable language it wouldn't be so fucking hard to parse the language. They could at least formalize things like order-of-operations regarding clauses and enumerated lists and give a better set of punctuation to work with. They should choose whether they want the language to be pure communications medium with a formal syntax or if they want it to be a completely flexible means of artistic expression full of nuances and hints that can only be understood by a sentient being who has studied the language in-depth for many years.
--
Want a free iPod?
Or try a free Nintendo DS, GC, PS2, Xbox. (you only need 4 referrals)
Wired article as proof
"I don't know much about US high school education, but I would expect that students coming into college would know how to write grammatical English"
The latter part of this statement makes me really believe the former.
"There are a dozen opinions on a matter until you know the truth. Then there is only one." - CS Lewis (paraprhase)
So, what we have here is somebody just saying, in essense, "Gee, Microsoft, why isn't your software at human-level AI? I mean, how hard can that be?" and is so utterly incompetent at assessing how hard grammar checking is that they are utterly unaware of how incompetent they are. (Hmmm, that sounds familiar, though this isn't quite the same.)
I invite Associate Professor of Marketing and E-Commerce Sandeep Krishnamurthy to try his hand at the AI problems he is upset that Microsoft hasn't waved a magic wand and fixed, though I feel obligated to warn him that as an associate professor of marketing, he's likely to be in for a world of intellectual hurt unless he's got some other source of knowledge and skill squirreled away somewhere, like a PhD in Computer Science he is for some reason forgetting to mention.... Perhaps then he would have some understanding of why even the mighty Microsoft has not yet produced the Perfect Grammar Checker....
On that note, check in with actual Linguists on the feasibility of the idea of a Perfect Grammar, too. You probably have a lot to learn there, too.
Hulk work hard on Grammar Checker for Microsoft! Program many long hours. Very hard to type with huge green hands and puny little keys! Many times get angry and smash keyboard. Many keyboards broken. Hulk also get help with grammar from Yoda. Yoda very wise. Maybe not best work in world, but Hulk take pride in work. Why puny University of Washington professor criticize hard work of Hulk? Criticism hurt Hulk's feelings. Hulk angry! HULK SMASH!
That should be "if I were a professor". It's the subjunctive mood. Betcha you wish you had a better grammar checker now!
"Would I use an intelligent grammar check? Yes, by all means. [... goes on to make suggestions ...]"
What you're missing is the fact that this is one of the hardest problems ever tackled by computer science. Not only that, but even a moderate improvement over what MS does now would likely require an order of magnitude more code and run-time computation, making it inappropriate for most usage!
MS Word does an OK job of spotting the most common errors, but if you're better at it than MS Word is, just shut the thing off. There, no problems.
As far as writing something that you KNOW is incorrect... ok, so you get a green line under text that you already know is a problem, but you don't intend to change. No big deal. Why is this an imposition?
OK. It's stupid.
There. Do I win?
That is all.
There are two open source grammar checkers available (Language tool and Queequeg ). Both have strengths and weaknesses and could do with a lot more work to improve on the number strengths they have. Unfortunately people with the necessary expertise rarely have the time needed to get involved on these kinds of projects. If anybody is interested and has some expertise then send me some email. Perhaps we produce something better.
The only things it flagged were "all your base are" (suggested "base is" or "bases are") and "for great justice" (sentence fragment).
Grammatical and sociological implications are left as an exercise to the reader.