How Would You Test a Spell Checker?

It’s become a bit of a running joke now on Slack when peoples spell checkers start to act a bit funny.

Some examples:
@g33klady got this on MS Teams

I got something similar to Hilary in Google docs, a red underline under a word with correct spelling that had " at the end. When I clicked to see what the issue was, it told me there were no suggestions but the red underline stayed :thinking:

@brimcrob also got this on Wednesday
image (1)

And of course this gem from @qualityfrog

So for a bit of fun, how would you approach testing a spell checker?


Instead of my usual “write an approach” tactic, just for fun, I present you with:

Buffalo buffalo Buffalo buffalo buffalo buffalo Buffalo buffalo.

This is a grammatically correct sentence. In word, it says all of the buffalo except the first one are incorrect. So it might be a good test.

Another one which annoys me, but I’m not going to do anything about it is that this box is mostly underlined as wrong.

Edit: Missed a context point there, but “this box” means the post-editing box on the club… thus here.


I could probably fix it, but it would take more than 2 seconds, so as an average user, why bother?

1 Like

Almost as if your browser dictionary is configured wrong Brian, which is probably not a bad approach route to take at all.

I would simply be not taxing my noggin, but rather exploring ways a user might find the spell checker an annoyance; finding ways that it might miss-configure; and find arguments to support red squiggles being a poor colour choice.

1 Like

At first thought, it seems like testing a spell checker might be easier to demonstrate that it has problems, rather than to show its ‘correctness’. There are just so many different things that would need to be tried with a spell checker, including…

  • Basic word spelling
  • Grammar
  • Recognition of pronouns
  • How does it know someone didn’t name their kid Aberaham Lyncoln?
  • If it correctly recognizes misspellings, what does it suggest instead? (This editor originally flagged misspellings, but now it doesn’t. What’s up with that?)
  • Does it ignore strings of numbers and/or characters, like 08/31/2020 or (2+2)*3=12?
  • Does it care about extra spaces in sentences?
  • Who about suggesting How when typing who?
  • What about it’s detection of incorrect use of its vs. it’s?
  • Are abbreviations OK?

Aside from targeted checks like the above, I might try pasting in large chunks of text (even from public domain books) to see what gets flagged - and compare against something like Word. A flawed baseline is still a baseline, right?


I think spell check is feeling a bit Friday today :thinking:

Screenshot 2020-09-18 at 09.36.02