Author Topic: online grammar checkers and their inability to spot errors  (Read 113 times)

Offline dnice

  • New Linguist
  • *
  • Posts: 1
online grammar checkers and their inability to spot errors
« on: February 06, 2018, 12:24:46 PM »
Good morning. I'm a college English prof with quite a few international students.  Our school uses, and encourages students to utilize its grammar checking component to help them with their writing.  However, there are quite a few mistakes that it doesn't seem to catch.  In thinking about it, I wondered whether the errors might not be at the syntactic level but rather at the practically expressed semantic level.  Here's a sample of a sentence that and other online grammar checkers found error-free, but which, to a native speaker, contains multiple errors.

Your help to improve me in English will remember in the long run.

My question is, why is this sentence technically correct?

Offline Daniel

  • Administrator
  • Experienced Linguist
  • *****
  • Posts: 1650
  • Country: us
    • English
Re: online grammar checkers and their inability to spot errors
« Reply #1 on: February 06, 2018, 06:18:41 PM »
Of course the sentence is not correct. Grammar checking software is just not (and probably will never be) as good as human proofreaders. Software like that usually relies on looking for common errors, especially types that are easy for a computer to find, like when a subject and verb (usually next to each other) don't agree in form (plural vs. singular). The sentence you gave above is not that type. On the one hand the specific errors are not common enough that the program is designed to look for them (the computer can't take really messing writing and fix it-- it's just like spell check for grammar and will polish a bit with near matches), and on the other hand you're right those examples are more about meaning (and forms required for specific verbs) than the sort of general structural errors the software identifies.

Technical details, guess about that sentence in particular: those words generally look like they might belong together when you only look at about 3 of them in a row. Many grammar checkers look at "n-grams" which are sequences of N words [often 3: trigrams] to see if they plausibly follow each other. When you look at parts of that sentence they don't look too wrong. But when you look at the meaning of the whole sentence and the form that is supposed to express it, the sentence falls apart. Like your students who are trying to put words together, the software sees those words and doesn't notice any immediate problems, but doesn't actually have the ability to proofread the final sentence as a whole. And your students feel like they've done the work of putting the words together and then hope that the software will fix it for them. Your advice is giving them a false sense of confidence and also telling them to ignore their own instincts.

In short, don't rely on grammar checkers without proofreading too. Especially for someone who doesn't know if the output is right or wrong. They're reasonable as reminders to check writing but not as the final word on whether something is correct, especially for learners who don't know how to verify it themselves. To be direct: I don't agree with your advice, and it will inevitably lead to this situation when students outsource their own learning/thinking/writing to a computer. Use it if you want to catch careless errors like typos but not quality control of writing in general. That won't work.

Most importantly this will work better for native speakers who have the experience to evaluate the correctness of such suggestions and fill in the rest themselves. It's like using google translate to translate text automatically: it works much better if you proofread it for meaning and form afterwards, in a language you know well (e.g. into English). Not so well if you translate blindly into another language you don't know and can't proofread it. If you wouldn't ask your students to use google translate for writing then probably also don't suggest grammar checkers.

Or for another comparison it's like auto-fill suggestions while typing on your phone. Sometimes right, often wrong. No substitute for proofreading.

Finally, I'm the sort of student who would have objected to this in class if you had suggested it. I'm very careful about where I put commas in my writing (and so forth). Even if a teacher proofreads my work by hand (and especially a computer!), I will not agree with all suggested revisions, often because I specifically intended to use a certain form, maybe a comma to add a pause rather than a more common usage. What I think teachers should really do is try to get students engaged more in their own usage, and try to promote meta-linguistic awareness: ask them why they wrote something a certain way. Was it a careless typo? (Easy to fix, and also not that important in the end.) Or was it because they didn't think about the alternatives? Did they not have a reason at all? By relying on grammar checkers they'll think less, not more. You could ask an exercise have them actually respond to the suggestions of the grammar checker and decide whether they agree or not, and why. (But I'm not sure that would be a good use of time, because grammar checkers just aren't that sophisticated, but maybe. Or some other active engagement exercise.) Consider the same situation with spell checkers: it has been said, and is certainly true to some degree, that people are lazier now about spelling because the computer can just do it for us. (And sometimes the suggestions are wrong! It's never perfect, though much better than grammar checkers ever will be.) So the best thing you can do as an instructor is to get your students to really think about writing. If you get them to think about where they put commas and why, you've succeeded. The actual usage should also not be thought of as "correct" or "incorrect", but just a choice, as long as they are doing it intentionally, and for effect.

For English learners there may be completely "incorrect" errors in the sense that they don't really know the target [native speaker English] because they're still learning, so that's a slightly different case. But still they should be engaged, and probably are more than your native speaker students, in trying to figure out why this or why that, and grammar checkers will actually, as I said, allow them to not think about this, and remember: it's hard for them. So it's comfortable not thinking about it. Really not the way to approach this problem. Much better would be correcting their own work, annotating errors they find: that's learning. Blindly using a grammar checker is not only often wrong, but also pointless: at the end of the day it doesn't matter if they've written a "correct" sentence in an essay they submit to your for class that neither of you will ever look at again in the future. What matters is what they learn from the experience.
« Last Edit: February 06, 2018, 06:40:10 PM by Daniel »
Welcome to Linguist Forum! If you have any questions, please ask.