Recipe: Product title matching

The challenge

Product titles vary in word order, extra descriptors, punctuation, and missing or added model details:

"iPhone 15 Pro Max 256GB"  vs  "Apple iPhone 15 Pro Max"  → same product
"Peugeot 208 Active"       vs  "208 Peugeot"              → same
"Samsung Galaxy S24 Ultra" vs  "Galaxy S24 Ultra 256Go"   → same

Recommended configuration

import { rate, findBestMatch, filterMatches, extract, createMatcher } from 'guess-rater'

const config = {
  algorithm: 'tokenSet',    // tolerates extra/missing words
  normalize: {
    removePunctuation: true,
    removeAccents: true,
    removeWords: ['the', 'a', 'an', 'le', 'la', 'les']
  },
  threshold: 78
}

Why tokenSet? Product catalogs often have extra words (brand names, capacities, colors). tokenSet compares shared tokens and ignores extras.

Why 78? Lower than person names because product titles naturally have more variation between legitimate matches.

Usage

// Find best matching product
findBestMatch('iphone 15', [
  'Samsung Galaxy S24',
  'Apple iPhone 15 Pro',
  'iPhone 15 Plus 128GB'
], config)
// { value: 'Apple iPhone 15 Pro', score: 91, index: 1 }

// Get all good matches above threshold
filterMatches('peugeot 208', [
  '208 Peugeot Active',
  'Peugeot 308 GTi',
  'Renault Clio',
  'Peugeot 208 Style'
], config)
// ['208 Peugeot Active', 'Peugeot 208 Style']

// Extract top 3 results for a search bar
extract(userQuery, productCatalog, { ...config, limit: 3 })

Handling word-order variations

Use tokenSort if your titles differ mainly in word order:

rate('Peugeot 208 Active', '208 Active Peugeot', { algorithm: 'tokenSort' }) // 100

Use hybrid when you have both word-order and extra-word issues:

const config = {
  algorithm: 'hybrid',
  hybrid: { tokenSort: 0.4, tokenSet: 0.6 }
}

Building a search function

const matcher = createMatcher({
  algorithm: 'tokenSet',
  normalize: { removePunctuation: true, removeAccents: true },
  threshold: 75
})

function searchProducts(query, catalog) {
  return matcher.extract(query, catalog, { limit: 5, return: 'entries' })
}

searchProducts('iphone 15', productCatalog)
// [{ value: 'iPhone 15 Pro', score: 91, index: 4 }, ...]

→ Algorithms guide | List helpers