Skip to main content

Generate random words based on Markov chains

Project description

Gibi is a random word generator, based on Markov chains.

It analyzes a text in order to deduce the probability of transition from one character to another, and then generates a completely random word that will look alike what you have feeded it as input.

Build status

develop:

build-develop

master:

build-master

Usage

Gibi can be used either as a CLI tool, either as a library.

CLI

A tool named gibi helps you to generate a matrix file, and then use it to generate words. It is self-documented using the –help switch, however here is a typical workflow.

$ gibi analyze corpus.txt matrix.gibi
$ gibi generate matrix.gibi

This will analyze the corpus.txt file into the matrix.gibi file, and then produce 10 random words.

API

Basic usage

In its simplest form, using the API looks like:

with codecs.open(path, 'r', encoding='utf-8') as f:
    n = FrenchNormalizer(f)
    m = Matrix()

    m.feed(n)

print(m.make_word())

You can also make something deterministic by providing a seed to make_word(). Given the same Matrix and the same seed, you will always get the same result. The provided seed is anything that Python’s random would accept. See the following example:

assert m.make_word(42) == m.make_word(42)

Using matrix files

You will probably pre-generate a matrix file using gibi analyze, and then load the resulting file like this

m = Matrix()
with open('matrix.gibi', 'rb') as f:
    m.load(f)

print(m.make_word())

This is much more performant, as generating the matrix can be time-consuming if the corpus is big.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

gibi-0.2.1.tar.gz (6.6 kB view hashes)

Uploaded Source

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page