suffix-thumb

find the optimal transormations between sets of words

npm install suffix-thumb

---

suffix-thumb tries to discover the way two sets of words map to one another, according to changes in their suffix.

It was built to learn rules about verb conjugations, but in a way, it is just a generic compression algorithm.

The assumption is that a word's suffix is the most changeable part of a word.

import { find, convert } from 'suffix-thumb'

const pairs = [
  ['walk', 'walked'],
  ['talk', 'talked'],
  ['go', 'went'],
]
let model = find(pairs)
/* { rules: [ ['alk', 'alked'] ],
    exceptions: {go:'went'},
    coverage: 0.66,
}*/

const pairs = [
  ['aail', 'aael'],
  ['bbil', 'bbel'],
  ['cil', 'cel'],
  ['snafoo', 'snabar'],
  ['poofoo', 'poobar'],
]
let model = find(pairs)
/*
  { 
    rules: [ 
      [ 'foo', 'bar', 2 ],
      [ 'il', 'el', 3 ]
    ],
    exceptions: {},
    percent: 100 
  }
*/

let out = convert('snafoo', model)
// 'snabar'

How it works

For each word-pair, it generates all n-suffixes of the left, and n-suffixes of the right.

any pattern between the two sets of words begins to pop out.

it reduces any redundancies in this list.

it then runs the patters on the dataset, to get a score, and any exceptions.

There may be wordlists with no helpful patterns.

Ideally, you should be able to take a list of word-pairs, create a model for them, and then delete the 2nd half of the word pairs.

MIT