Grammar checkers and sand in the carpet

This Tech Talk column originally appeared in The Exponent (University of Alabama in Huntsville), Vol. 32, No. 14 (December 7, 2000). Digitized issue: UAH LOUIS archive. Reproduced here courtesy of that archive.

After fighting through an awful traffic, I finally returned home from the beach. I was in the mood for a music and hastily dropped in a CD. Beginning the unpacking process, I pulled a pants from a luggage to find it covered in sands. Disregarding the many advices my mom have given me, I shook the garment and ended up getting sands in the carpet and furnitures.

If the previous paragraph sounded strange, then you’re probably a native speaker of English. Ignoring the “What I Did on My Summer Vacation” flavor, its weird ring comes from the use of mass nouns as count nouns. Mass nouns denote undifferentiated collections or concepts whose members you wouldn’t or couldn’t ordinarily count. Mass nouns don’t take the plural form, so “sands,” “advices,” and “furnitures” are all ungrammatical. In addition, the indefinite articles “a” and “an” don’t precede mass nouns, so “an awful traffic,” “a music,” and “a pants” are ungrammatical.

In case you’re wondering, this is still the tech column and not the Grammar Gestapo Gathering. However, those who deal with math or computer science enough begin to pay much closer attention to language—not much of a surprise given that programming languages are popular tools with computer scientists. Mathematician Kurt Gödel, in the proof of his incompleteness theorem, invented a formal language that provided an avenue for the integers to “talk” about themselves and create an unprovable statement about the integers. We don’t want to leave out our friends in the liberal arts: philosophers and logicians use a formal language known as propositional logic to formulate claims, arguments, and proofs.

With that little digression out of the way, let’s get to the point. Grammar checkers in word processors are terrible. They routinely flag perfectly grammatical constructions and miss blatant errors. A classic example is the sentence “The horse raced past the barn fell,” which is grammatical (it’s a reduced relative clause: “The horse [that was] raced past the barn fell”) but which most grammar checkers will flag as a fragment or subject-verb disagreement. Meanwhile, something like “Me and him goes to the store” often slips right through.

Why are grammar checkers so bad? Natural language is hard. English is full of ambiguity, idioms, and constructions that depend on context and world knowledge. Writing a program that can reliably distinguish grammatical from ungrammatical English is, to put it mildly, a non-trivial problem—one that artificial intelligence researchers have been working on for decades. The checkers shipped with Microsoft Word and its competitors are heuristic toys, not serious linguistic engines.

So what’s a writer to do? Learn the rules. Read good writing. When in doubt, rewrite. And for heaven’s sake, don’t shake sandy clothes over the carpet.