Shouldn't use the word in the definition

It seems the only variables are the dictionaries you input.

Dictionaries don't change much over time.

Is there anyway to generate more fake words than what you have on your web site?

I am seeing most of the words generated are already registered for .com

A common feature of real dictionaries is to show the pronunciation of words, like

gin·ger·ly /ˈjinjərlē/

There's an open pronunciation dictionary (Carnegie Mellon University Pronouncing Dictionary) that could potentially be integrated to display how to pronounce the generated word!

Complete a whois look-up on each name generated on https://www.thisworddoesnotexist.com/ (with a referral code for for your revenue/hosting costs) to see if the .com is available for the generates name.

I believe the vast majority of these regexes are typos: https://github.com/turtlesoupy/this-word-does-not-exist/blob/master/title_maker_pro/bad_words.py#L16-L51

Those first / characters shouldn't be there.

This requires at least two function in transformers that don't seem to exist anymore. I see that at least one of them existed in Transformers 3.02. I tried installing both 3.02 and 3.5.1 (using pip) to see if I could get a compatible version. Both of them failed to install because of missing source files (Windows 11, VS 2022 Community).

So now that I've changed my whole OS to Windows 11, I can't run my favorite plugin for my bot that I made using this-word-does-not-exist. Very disappointed. Can you please update the code to work with all the latest versions of the requirements? Thank you.

https://l.thisworddoesnotexist.com/CvBh

The syllable section is treating the apostrophe as a pronounceable syllable (and missing the dot before the o)

Just a quick bug i thought I'd share. In the picture below, the word in the example sentence does not match the word generated:

Maybe it's a plural form? In which case that is a job well done 👍

EDIT: Here is the link to that word: https://www.thisworddoesnotexist.com/w/terezi/eyJ3IjogInRlcmV6aSIsICJkIjogInRoZSBmZW1hbGUgZ2VuaXRhbGlhIiwgInAiOiAibm91biIsICJlIjogImJhYmllcyB3aXRoIHRldGVyZXppIn0=.ePYuH6eRP7b7j1N_lFf7O03fPQ220Ydmx0c2vTkOlsM=

It currently adds a +proto which is in-spec but unsupported by google

from title_maker_pro.word_generator import WordGenerator
word_generator = WordGenerator(
  device="cpu",
  forward_model_path="forward-dictionary-model-v1",
  inverse_model_path="inverse-dictionary-model-v1",
  blacklist_path="blacklist.pickle",
  quantize=False,
)

# a word from scratch:
print(word_generator.generate_word())

I tried the above code and I get the following error:

m = start_end_re.match(word.misc)
TypeError: expected string or bytes-like object

I printed word.misc and it was "None". Any help for a fix is appreciated. Thanks in advance

Would it be useful to update this project with a better language model, or is that redundant?

Hey there, first of all: great work. I am getting the following error when running the sample code. Would someone be able to give a hint where i am doing wrong?

Best

0%| | 0/1 [00:00<?, ?it/s]Traceback (most recent call last): File ".../main.py", line 12, in <module> print(word_generator.generate_word()) File "...\title_maker_pro\word_generator.py", line 72, in generate_word use_custom_generate=True, File "...\title_maker_pro\datasets.py", line 638, in generate_words example_match_pos_pipeline, l_example, start_title_idx, len(t_rstrip) File "...\title_maker_pro\datasets.py", line 346, in approx_pos m = start_end_re.match(word.misc) TypeError: expected string or bytes-like object

Visiting https://thisworddoesnotexist.com/ results in a HTTPS error, as the certificate is only valid for www.thisworddoesnotexist.com and not thisworddoesnotexist.com.

https://l.thisworddoesnotexist.com/twJc

This word ends in a comma. I've only found this one occurrence but wonder if other characters might run into the same problem too.

Very fun project, by the way!

Maybe during sampling the percentage of matching regexp (correctness) + percentAnge of matching outside blacklist (creativity)

For human-inputted words, the "a word that does not exist" text should be changed to something like:

a word that might exist, it was input by a human and defined and used by a machine learning algorithm.

I was wondering if it was possible for thisworddoesnotexist.com to have an API that other developers can use so they don't have to set this project up themselves

Example URL: https://www.thisworddoesnotexist.com/w/jesdabest/eyJ3IjogIjxTQ1JJUFQ-YWxlcnQoJ2plc2RhYmVzdCcpPC9TQ1JJUFQ-IiwgImQiOiAiYW4gYWxlcnQgcGxhY2VkIGR1cmluZyBhIHZpZGVvIHJlY29yZGluZy4iLCAicCI6ICJub3VuIiwgImUiOiAiPFNDUklQVD5hbGVydCgnamVzZGFiZXN0Jyk8L1NDUklQVD4iLCAicyI6IFsiPFNDUklQVD5hbGVydCgiLCAiJyIsICJqZXMiLCAiZCIsICJhYmVzdCIsICInIiwgIik8L1NDUklQVD4iXX0=.53iUkLV3kFioMfmibqWu6403RBPA3xxYXfEiRheMoc8=

To avoid this, HTML entities (namely "<" and ">") should be escaped (to < and >) before being written to HTML.

You only have 40 characters to play with, but that would in principle be enough to load malicious code of arbitrary length from another URL. Of course, cross-site scripting on thisworddoesnotexist.com is of limited impact anyway!

Possible solution:

• Get Wikitext-102
• Use stanza to parse out words
• Anything occuring > threshold times added to dictionary
• Union all the article titles in Wikipedia (date stripped)
• Union words from source dictionary

IS: could not find the license
SHOULD: would like to know the license, would love if it is MIT licensed

When running word_generator.generate_word_from_definition we get a default progress bar.
Is there a way to disable this?

There are times I'd like to literally make a joke entry with my own word and specify the pronunciation, word type, and definition. Any way you could add the ability to put those in when doing "Write your own"?

A simple ajax refresh to reload the page would help users play with the website generating words to their hearts content. One way to circumvent quota limits would be to generate a static list of words (say, 10000 every other day) and use this list on the front-end instead of making direct calls.

Hi friends,

When executing it in my Anaconda prompt I got this error:

I tried:

1. redownloading the blacklist.pickle.gz
2. try with just "blacklist.pickle"
3. uncompress it
4. changing the path of the file
5. dive in stackoverflow for three days for a solution
6. using double slashes in the filepath
7. double slash + different filename

Finally I guess '\x1f' refers to an error reading '\b' in the blacklist_path but I don't really know how to solve it.

Any help please? 😕

Like this project a lot

Interested in:

1. Set a character length for the generated words
2. Ignore or generate specific prefix/suffix words

Can you point me to the code where I can build this out ?

help with SEO

Such as 'mraada'

Probably shouldn't have the bot try to reply to a thread that already has a definition in it. Should make sure people can use it within a reply thread (e.g. define covfefe)

Right now this gives a "something went wrong"

Is there any thing special thing about English or this can be trained and used for languages like Japanese, Arabic, Russian, Hebrew?

• Retrain on same dataset in a different auto-regressive order
• Maybe add in examples for fun

Examples use adjectives as nouns, etc.

Possible fix:

• Use stanza to POS tag examples
• Ensure POS tag matches what the definition was hoping for

Right now can get "stuck" at an ID -- probably just switch trying to compute 50 most recent after an id.