Why Spell Check Leaves Much to Be Desired

There is a myth pervading America, sabotaging the efforts of students and professionals alike. That myth is the myth that spell check and grammar check work well.  People everywhere are relying on them as their sole source of editing, much to the detriment of their writing. 

To understand just why this dependence is more a hindrance than a help, let’s look at how spell checkers and grammar checkers work.  This post will focus on spell check first because it’s more widely used than grammar check.

How Spell Check Works

Millions of people use their word processor’s built-in spell checker without knowing how it actually works.  It’s certainly not necessary to know what is going on behind the scenes in order to make use of this tool, but it can help you make an educated decision about the suggestions it gives you.

Because spell check is more widely used than grammar check, we’ll start with it first.  In simplest terms, spell check compares every word you type to a built-in dictionary; if it can’t find the word you type in its dictionary, it marks it incorrectly spelled (usually by underlining it in red).  While it’s nice to know what words are misspelled, simply identifying them for you doesn’t necessarily help you figure out how to fix the error.  At best, it means you have to pull out a dictionary yourself and search for the word.  The problem with that is you have to know how a word is spelled before you can look it up in a dictionary!

Spell checkers therefore have a second built-in step; after they identify an incorrectly spelled word, they provide suggestions for the correct spelling.  You simply scroll down the list of suggestions until you find the word you meant to type. Sounds helpful, right?

The main issue with spell check’s utility is the number of suggestions you receive.  Not counting synonyms, only one word should fit in the place of the word you spelled incorrectly.  How, then, is the computer generating a list of possible replacements? Doesn’t it know what you meant to write?

To understand spell check’s limitations a bit more, let’s break how it generates suggestions.  After identifying the incorrectly spelled word, the spell checker needs to figure out how it’s been misspelled.  To do that, it generates a list of words (both real and nonsense words) based on the different types of misspellings that can occur:

• Deletion: a letter is missing (for example, “stacing” instead of “stacking”)
• Insertion: there is an extra letter (for example, “eatting” instead of “eating”)
• Substitution: one letter replaces another (for example, “thwee” instead of “three”)
• Transposition: letters have switched places (for example, “teh” instead of “the”)

Here’s an example of a (very abridged) list of words a spell checker generates as it tries to figure out how you misspelled the word it’s trying to fix:

Misspelled Word:	siting (instead of "sitting")
Deletion:	saiting, sbiting, sciting, sditing, smiting, sitting
Insertion:	sting, siing, sitng, sitig, sitin
Substitution:	sating, siring, seting, sitang
Transposition:	isting, stiing, siitng, sitnig

 

From among the nonsense words generated, we have some real words that are not misspelled.  Using percentages gleaned from extensive analysis of a corpus of documents, the spell checker next ranks these words according to how likely it is the error occurred.  The final result is the list of suggestions the spell checker presents to you when you spell check your document.

The key idea to note here is that the spell checker does NOT know what word you meant to use.  It’s not actually taking into account the  meanings of the words and whether they fit into the context of the sentence.  The spell checker also doesn’t know if you made more than one type of error (or more than one instance of one type of error) when spelling your word.  While it will try to generate words based on more complex errors, it will rank more likely misspellings higher on the list, so the word you meant to type may be much farther down the list.

Why is this a limitation when it comes to spell checkers? We all know no one scrolls past the first few suggestions, if they bother to read the suggestions at all.  Most people just hit “change” automatically.  Because semantics aren’t integrated into the spell checker code, the words it suggests to you may not be the correct ones. 

Without getting into grammar check territory, there are some errors that spell check cannot find.  It can’t tell when you’ve used repetitions of a word incorrectly.  For example, “I’m very very sorry” may be acceptable, but “I’m writing about about the author” isn’t. It also can’t check whether or not your main points are in order or even if “Firstly” comes before “Secondly” in your paper! And let’s not even go into homonyms….

Spell check is definitely better than not editing your paper at all, but ONLY if you read the suggestions before clicking “accept” on the suggested change.  If you’re a non-native speaker of English, then you can still benefit from spell check without sacrificing accuracy; look up each of the suggested words in a dictionary until you find the word with the meaning that fits what you were trying to write. It will take more time, yes, but with repetition you will make fewer mistakes, learn more vocabulary, and even be able to spell better than many native English speakers!

If you’re interested in knowing the nitty gritty about how spell check is coded in a programming language, James Matthews wrote a great introduction called “How Does Spelling Check Work?”  I should warn you, though, it’s very technical. 

Examples of Spell Check Fail

Most people need to see something to believe it, especially when it comes to a tool many people wouldn’t be able to write a paper without. Here are examples of spell check failing in action:

• Taylor Mali has a special place in my heart.  He’s the kind of teacher I want my future children to have! Taylor Mali’s performance of his poem “The The Impotence of Proofreading” is classic and I have shown it to every high school and college class I could. He ran this poem through spell check and was told it was 100% correct.  Be warned, though, that there are a few sexually suggestive moments in the poem, but nothing too explicit. If you haven’t seen his “What Do Teachers Make” video, it’s phenomenal!
• Jerrold Zar’s poem “Candidate for a Pullet Surprise” is another poem that demonstrates just how wrong a spell check can be when it tells you your paper has no errors.  This poem is a classic! 
• There’s an entire web site dedicated to the iPhone’s auto correct feature.  Be warned, though, DamnYouAutoCorrect.com has many explicitly sexual entries.

Let the preceding links serve as cautionary tales for why it pays to be as smart as the tool you use.  Just like math teachers won’t let students use a calculator before they know how to add, subtract, multiply, and divide without it, you need to be able to spell before you rely on a spell check to catch errors for you.  Spell check should be your last defense and the errors it catches should be careless ones on your part that you immediately recognize when it points them out.

But don’t just take MY word for it.  Read Johanna Sorrentino’s article “Is Spell Check Creating a Generation of Dummies?” and Jill Baughman’s “25 Reasons Not to Trust Spell-Check When Job Hunting.”

Photo source: BureauCrash

Posted in: spelling