Pronounceable returns a score between 0 and 1 for how pronounceable a string is.

go get github.com/felixdorn/pronounceable

A dataset is just a list of words separated by newlines.

We provide a few datasets of various size.

• 10k words
• 50k words
• 100k words (recommended)
• 333k words (not recommended)

You may download them to include in your project.

• score > 0.6: good
• 0.5 > score < 0.6: meh, it depends
• score < 0.5: terrible

Be aware that the larger the dataset, the higher the minimum score. See how it works.

package main

import "github.com/felixdorn/pronounceable"

dataset, _ := pronounceable.NewDatasetFromFile("wordlist.txt")

dataset.Score("incomprehensibilities") // ~0.66 with 100k words dataset
dataset.Score("hello") // ~0.79 with 100k words dataset

Given the following dataset:

hello
world

We generate mono-grams, bi-grams and tri-grams for each word.

We end up with a map that looks like this:

monograms:
  h: 1
  e: 1
  l: 3
  o: 2
  w: 1
  r: 1
  d: 1
bigrams:
  he: 1
  el: 1
  ll: 1
  lo: 1
  wo: 1
  or: 1
  rl: 1
  ld: 1
trigrams:
  hel: 1
  ell: 1
  llo: 1
  wor: 1
  orl: 1
  rld: 1

Then for a given word, let's say chicken, we generate its mono-grams, bi-grams and tri-grams.

monograms:
  c: 2
  h: 1
  i: 1
  k: 1
  e: 1
  n: 1
bigrams:
  ch: 1
  hi: 1
  ik: 1
  ke: 1
  en: 1
trigrams:
  chi: 1
  hic: 1
  ick: 1
  cke: 1
  ken: 1

For each of these n-grams, we compute their score.

log(dataset[n][ngram] / len(dataset[n])) * (5+n)

Where n is the n-gram length and dataset[n] is the map of n-grams.

We multiply the score by 5 + n to reward longer n-grams proportionally to their length.

We can then calculate the score for a given word by summing the scores for each n-gram.

sum(scores) / 1.5 * len(word)

We multiply the score by 1.5 to penalize longer words.

Pronounceable was created by @afelixdorn under the MIT license.