A simple procedural text generator implementing Markov chains
npm install markov-text
var Markov = require('markov-text');
const trainingText = 'Lorem ipsum dolor sit ammet'
options = {...}
const loremGenerator = new Markov(options) // Setup generator
loremGenerator.seed(trainingText) // Seed chain with "training" text
const generatedText = loremGenerator.generate(5) // Set length of the generated output.| Method | Arguments | Returns | Description |
|---|---|---|---|
| seed | trainingText | - | Seeds the generator with the training text. The generator splits the text into nGrams (pieces of n-characters or n-words length, depending on the mode)analog to chainlinks. |
| generate | outputLength | output | Returns the generated text of the specified length. (Length is in ngrams not in charcters/words) |
You can pass in an options object when instancing the generator that accepts the following options:
| Property | Type | Options | Default | Description |
|---|---|---|---|---|
| order | integer | 1-n | no | Specifies the length of the ngrams (chainlinks). Longer chainlinks will produce more coherent text but less 'creativity' |
| mode | string | 'single' | 'multiple' | 'multiple' |
You can view the chain building step by setting up the enviroment variable DEBUG_CHAIN to true
DEBUG_CHAIN=true node myTextGenerator
Provided are two examples you can run using:
Uses a excerpt of Franz Kafka's Metamorphosis to generate sentences of the desired length.
npm run example:metamorphosis
Uses all the names from characters of LOTR to generate a new one
npm run example:lotr
Copyright for training texts is owned by their respetive authors and is not protected by the license of this library
npm test
This is mainly my best shot at implementating in Javascript what is explained in this great series by starbeamrainbowlabs