The G Files

The G Files Part 3: 10 Failures of Common Sense

As I said in The G Files Part 1, some of the mistakes made by grammar-checking software make syntactic sense but are meaningless or nonsensical in the real world – like Chomsky’s famous “Colourless green ideas sleep furiously” sentence. This is because the software knows which words can be nouns, verbs, and so on; but it doesn’t have any semantic knowledge of the real world, or any “common sense”, so it doesn’t know which nouns can sensibly do which verbs.

Here are 10 examples I found entertaining. (Note that whenever I say “the software thinks…”, I mean something like “the syntax tree represented in its data structures corresponds to…”)

1. There’s nothing like a survey to keep a group focused

“Surveys, focus groups and interviews are examples of primary research methods.” Here, the software complains about a comma between subject and verb. (This is the kind of error you’d make if you wrote “The cat, sat on the mat.”) Presumably it has parsed “surveys focus groups” in the same way as “lenses focus light-rays”, with “focus” being a verb and “groups” being the things that get focused. This is an example where an unconfident writer might take the software’s advice and remove the comma.

2. Low-interest rates must be the ones no one signs up for

The software wants to change “low interest rates” and “high interest rates” to “low-interest rates” and “high-interest rates”. You need genuine, external, semantic knowledge to understand the difference between high interest rates and high-interest loans, or between low interest rates and low-energy light bulbs.

3. You have to know the drug was taken

The software wanted to change “When treating an overdose, it is important to know the drug taken” to “…to know the drug was taken”. Well, yes, it’s important to know that too, but…

4. The company just filled in the survey itself anyway

“The company asked its employees to try the new product, then complete a survey about it.” The software wanted to change “complete” to “completed”. You need fairly subtle human semantic understanding to realise that the company didn’t complete a survey, it asked the employees to complete one.

5. The amazing multitasking catalyst

“The catalyst speeds up the forward and reverse reactions equally.” The software wanted to change “reverse” to “reverses” – I guess it thinks the catalyst speeds up “the forward”, whatever that is (perhaps a centre forward in football?), and the other thing the catalyst does is that it reverses reactions equally.

6. Regulatory creep?

In “Regulations may quickly change for the business”, the software objected to “for” as a redundant preposition. I suppose regulations may change the business as well, as an indirect effect; but that wasn’t what the writer intended.

7. Spare some land, guv?

“Workers may be made redundant and spare land and goods can be reallocated or sold.” The software suggested inserting a comma after “land”. Presumably its interpretation is that workers may be made redundant and workers also may spare some land?

8. Look out for those molecule increases!

In “As the molar mass of a molecule increases”, the software underlined “a molecule increases” and said “indefinite article with plural noun”, meaning it would prefer “As the molar mass of molecule increases”. I guess the “plural noun” means it thinks “molecule increases” are a thing?

9. Oh no! I’m 14.3%!

“The scientists determined the percentage uncertainty in the activation energy of a reaction to be 14.3%.” The software thinks it should say “…to being 14.3%”. It took me a while to figure out what was going on here. It thinks we’re talking about “a reaction to being 14.3%”, like “a reaction to being fired”. And then the scientists calculated the uncertainty in the activation energy of that.

10. You can take the asteroid out of space, but you can’t take the space out of the asteroid

Finally, a related error: the software is quite good at noticing when a pair of adjacent words could be joined to make a new word (and therefore the space between them could be an error), but it’s terrible at figuring out whether the separate or combined word would fit better in the context. For example, it keeps wanting to change “a steroid” to “asteroid”.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s