voweler

goal: insert vowels in text that has no vowels (unless word starts with vowel)

idea:

• precompute double metaphones for dictionary words
• precompute voweless versions of dictionary words
• at runtime, compute double metaphone for input word
• if double metaphone is null (e.g. metaphone('w')), get metaphone for word + 'a'
• filter dictionary by words whose voweless, sorted representation match input word (sorted)
• compute a score defined as the mean levenshtein distance between primary and secondary metaphones for each dictionary word against input word
  • that is, for each dictionary word's metaphones [dp, ds], and the input word's metaphones [ip, is], compute (leven(dp, ip) + leven(ds, is)) / 2
• filter the wordlist by smallest possible score and grab the top

using a wordlist that is sorted by usage and is relatively small (20k words) shows best results.

status: isn't horrible

use

$ npx voweler hppy brthdy
happy birthday

or import to your project (why?)

const voweler = require('voweler')

const vowelized = voweler('a sntnc wtht vwls')
// a contains without wolves
// -_- -_- -_- -_- -_- -_-

License

MIT