fictionary 0.1.0

Generate made-up words following the patterns used by real English words.

Generate made-up words following the patterns used by real English words.

Using Fictionary

Install with:

pip install --upgrade fictionary

You can learn how to use fictionary as a command-line tool by running fictionary -h:

usage: fictionary [-h] [-v] [-c COUNT] [-m LENGTH] [-x LENGTH]
                  [-d {all,american,british}]

A made-up word factory, following standard English word rules.

optional arguments:
  -h, --help            show this help message and exit
  -v, --verbose         Be verbose.
  -c COUNT, --count COUNT
                        The number of words to create.
  -m LENGTH, --min-length LENGTH
                        Only make_model words of LENGTH chars or longer.
  -x LENGTH, --max-length LENGTH
                        Only make_model words of LENGTH chars or shorter.
  -d {all,american,british}, --dictionary {all,american,british}
                        The dictionary rules to follow: american, british, or
                        all

Running it looks a little like this:

$ fictionary
nivenver

$ fictionary -c 4
cest
colped
burpen
flumat

Library Usage

And you can also use it as a library:

>>> import fictionary

>>> fictionary.word()
'regagreagised'

And if you want to create your own models:

# Create a model and add a couple of words to it:
m = fictionary.Model()
m.feed('table')
m.feed('babel')

# Now we can generate words!
# (This model is capable of only 2 fictional words)
print(m.random_word(5, 5)) # tabel
print(m.random_word(5, 5)) # bable

# If you're building a model with *lots* of words, generating the model
# can be slow, so you can save the generated model to a json file:
with open('my-fictionary-dict.json', 'w', encoding='utf-8') as fp:
    m.write(fp)

# And then later you'll want to read it in with this:
# (You still need to supply a list of 'real' words, for collision detection)
new_model = fictionary.Model(words=['table', 'babel'])
with open('my-fictionary-dict.json', 'r', encoding='utf-8') as fp:
    new_model.read(fp)
print(m.random_word(5, 5)) # bable

Why???

Why not? It is particularly good for generating memorable yet reasonable length passwords, although I’m not sure how secure those passwords would be given that they follow well-defined patterns. One day I might sit down and work it out.

How it Works

When it runs, fictionary loads a data structure called a Markov chain, which represents the patterns of letters found in the words in the dictionary (e.g. The most common first-letter is ‘s’. The most common letter following ‘s’ at the start of a word is ‘t’ etc.). Fictionary is supplied with 3 models out of the box:

Model	Description
all	Includes all words is both british and american wordlists.
american	Includes English words, using American spelling.
british	Includes English words, using British spelling.

Once fictionary understands the patterns of letters used in words in the English language, it can use these rules to generate new, nonsense words that look like English words, but aren’t. It’s so easy for the Markov chain to accidentally generate a real English word that we have to check each generated word against a dictionary to make sure it isn’t.

Releasing

These are notes for me, as is probably obvious:

• Check the README

• bumpversion

• python setup.py sdist bdist_wheel

• twine upload dist/*.*

To Do

The following is my to-do list for this project:

Allow Valid Words

Add a flag to turn off ‘real-word’ validation.

Word Generation Rollback

Rejecting words that are too long or short is reasonably expensive. I may refactor this to rollback and remake choices until a valid ‘word’ is reached. Or I may find something better to do with my time.

Optimize Long Words

Make word-generator bail out as soon as max-length is encountered.