Comment by bisRepetita
6 years ago
Great stuff. Found a real word:
refactoring
the systematic and systematic reworking of a piece of text to reduce unnecessary redundancies
Interesting that the made-up definition is pretty much the real one.
https://www.thisworddoesnotexist.com/w/refactoring/eyJ3IjogI...
Got a few comments about that one and a few others! I just updated the existing word blacklist and re-deployed.
"fibroblastosis" -- appears to feature in some medical journals
"bryosphere" -- something to do with moss
"biosprint" -- a brand of yeast.
Many of these 'non-words' have already been taken....
Oh, that second one is cute. I read up on bryophyes (moss and friends) for an exceedingly brief stint back in my undergrad days. Pronunciation similarity to bio made for many "bryo" puns.
I got Undercrowded
Oh, I'm surprised the "blacklist" isn't just the standard English dictionary. I'm sure I'm just being naive though. Why not just blacklist any word that already exists in English?
There are some subtleties (e.g. hyphens, derived forms, bigrams, etc.) but the biggest problem is that most English dictionaries don't have entries for every scientific word / piece of internet slang. I ended up tokenizing Wikipedia for a blacklist and still missed a lot :(
5 replies →
That may be because there isn't a single definitive list of all words in the English language
https://www.lexico.com/explore/how-many-words-are-there-in-t...
If you're using a blacklist, it it really machine learning? Or are you using it to re-train?
blacklist is probably to avoid cases where it randomly generates a real word like above two cases, so that blacklist filter is probably applied after the ml stuff.
1 reply →
Data scientist here. It's common to define boundaries for a machine learning algorithm by hand. Think of telling a chess AI that it can't move pieces off the board.
Unspooled and Hardstyle popped up for me. Perhaps you should do a google search for generated words before displaying them to prevent existing words from being shown.
Or maybe the ML algorithm behind is just using an existing dictionary and performs no operation.
1 reply →
It gave me "Intermodulate" a minute ago. https://en.wikipedia.org/wiki/Intermodulation
That ones a little more fuzzy; intermodulate doesn't occur very much in discourse (e.g. not in the wiki article at all) even though it would naturally be related
3 replies →
I got "cyberpolice" which seems to be a real word.
I also got "deflategate" which i love - it must be an upcoming scandal! :-)
Deflategate was a National Football League (NFL) controversy involving the allegation that New England Patriots quarterback Tom Brady ordered the deliberate deflation of footballs used in the Patriots' victory against the Indianapolis Colts in the 2014 American Football Conference (AFC) Championship Game.
https://en.wikipedia.org/wiki/Deflategate
> it must be an upcoming scandal!
It was, in 2016.
https://en.wikipedia.org/wiki/Deflategate
Multiple people managed to find it? How likely is it to generate the same non-word more than once? Is there a limited set?
I also got «invention» up as a result.
I just saw "gofundme".
I got "pataphysical".