Can you please remove misspelt words from your AgileWords.txt dictionary?

jimcheetham

Memorable Words are a great way to communicate between humans, but if one of the words generated is actually misspelt, it will cause confusion and failure.

I generated a password that included the word 'wierd', and copy/pasted it into use. I typed it into my phone to send out-of-band to a supplier ... where it got autocorrected, and I didn't notice, and of course they couldn't use the word they received successfully.

This problem would also have happened if I'd read the passphrase out over the phone, if I didn't notice the misspelling myself.

In AgileWords.txt, you have the correct 'weird' and 'weirdo', and also the incorrect 'wierd' ... you might have other misspellings in there too, and I think you should try to eradicate them

---

1Password Version: 7.6
Extension Version: 70600005
OS Version: macOS 10.15.6
Sync Type: Not Provided
Referrer: forum-search:dictionary misspellings

Ben

Hey @jimcheetham

I think that's a fair point. :) I'll bring it up with our security team and see if that is something we can make happen. Decreasing the size of the wordlist does have negative side effects, but it may be worth while in this case. If you spot any other words that are misspelled or objectionable please let us know.

Ben

ref: dev/core/core#2347

jimcheetham

I agree you can't sensibly decrease the total dictionary side without reducing entropy and also potentially encountering alignment issues with whatever data types you're using for the indexing ... but if the problem isn't too big you could probably replace the problematic ones with alternatives.

Without access to a full dictionary API checking your wordlist would be difficult :-) I've also realised that you have some archaic words in there that might not be recognisable, and potentially homophones/heterographs might cause problems when being communicated in spoken form, but there's only so much you can reasonably guard against.

However, within the resources I have available, I started off by removing everything that was in macOS's /usr/share/dict/words, leaving me with about 1000 candidate words. I eyeballed that down to 125 (some trepidation about US/UK spelling alternatives), then checked them in Wikitionary - if I didn't think the definition in there cleared up the word for me (after all, we don't all have to know what a word means in order to spell it/use it) then I kept it for the final list of 39 entries ...

These appear to be just plain wrong :-

abscound
aggeus
bildad
boccie
bowdoin
chieg
desicate
englex
fracus
frusta
haag
impotant
josue
manasses
mhammad
micheas
morcar
osee
stigand
venae
verandah

These are such old forms that they're arguably wrong / at least misleading, and probably have alternative homonyms in much more common usage :-

bowlder
cocoanut
compleat
carcase
diseuse
gayety
traumata
shlemiel
skurry
somalo
wierd

These are acronyms :-

fdic
naacp
fdic

These are too technical for most people ... even though I like them ... :-

vlsi
ktext
emacs
unix
vaxen

MerryBit

These appear to be just plain wrong :-

English is not my first language, so I tried to search for some of those words with Google, and several of them turned up sensible results. For example, "boccie" is an Italian game, Stigand was the name of the Archbishop of Canterbury in the 11th century, and frusta is both a whip and the plural form of frustum, a geometric shape. Verandah is also the correct spelling of a roofed porch, according to several dictionaries, though I've never seen it spelled that way myself.

So, many of the words may be obscure and archaic, yes, but not plain wrong. Given the enormous amount of words in the English language though, they should not be too difficult to replace, one would think. :)

vplewis

@jimcheetham @MerryBit Since we're doing this, let's add cocoanut to the Not Misspelled List. and of course a certain Izaak Walton would disagree that compleat is incorrect. And thanks for the " /usr/share/dict/words" link. Never knew that existed.

jimcheetham

The first check, are words 'real', has to agree on some commonality of knowledge. The point of a Memorable Word is that it has to be meaningfully remembered by a human as something better than 'just a string of letters', and from that perspective unfamiliar/obscure words are just fine.

Words that have multiple spellings/variants, especially considering American vs English usage but also 'archaic', are a touch more difficult. Chili-with-one-ell vs Chilli-with-two-ells for example. They probably take only slightly more mental space to store/process though, and as such are probably fine as well.

However, remembering words is only one use-case for these things. I also use them for communication between people, for shared secrets. That's when they cease to be Memorable Words, and perhaps try to become Unambiguous Words For Communication. My problem was caused by the iOS dictionary not including 'wierd' (even though it's "a word") and there's probably already a conflict in iOS between 'color' and 'colour' based on my Location settings too. That adds to the (tiny) problem of "just plain wrong" entries identified above. The recipient has to recognise the word I use and be able to spell it, which in a spoken conversation means they can simply ask me ... not so easy in a recorded message, but I could always make an effort to spell out any words I anticipate as being unfamiliar.

It might be that this new use-case is very different indeed, and perhaps deserves a totally different dictionary with far fewer words in it - which overcomplicates the user interface for many users I suspect.

Anyway, I hope I've ben able to identify the type and scale of the 'problem' I had, and 1Password have now noticed it at least :-)