Pular para o conteúdo principal

Stemmer

The Stemmer class reduces a Portuguese word to its stem using the RSLP algorithm (Orengo, V.M. & Huyck, C. (2001), A Stemming Algorithm for the Portuguese Language).

It complements Metaphone rather than replacing it: Metaphone normalises how a word sounds, the stemmer normalises its morphology, so inflections of the same word share one stem.

Basic Usage​

use ByJG\WordProcess\Portuguese\Stemmer;

$stemmer = new Stemmer();
echo $stemmer->stem('casas');     // cas
echo $stemmer->stem('correndo');  // corr
echo $stemmer->stem('correram');  // corr
echo $stemmer->stem('falou');     // fal

The input is trimmed and lowercased (mb_strtolower) before the rules are applied.

How It Works​

The steps are applied in this order; within a step, the first matching suffix wins:

  1. Plural reduction
  2. Feminine reduction
  3. Adverb reduction
  4. Augmentative/diminutive reduction
  5. Noun suffix reduction
  6. Verb suffix reduction (only when step 5 made no change)
  7. Vowel removal

Customizing​

The constructor accepts two optional arrays, so real corpora can be handled without editing the rule tables:

$stemmer = new Stemmer(
    protectedWords: ['solar', 'neural', 'docker'],
    extraExceptions: ['ar' => ['nuclear', 'muscular']]
);

echo $stemmer->stem('solar');    // solar  (never stemmed)
echo $stemmer->stem('nuclear');  // nuclear (the "ar" rule skips it)
echo $stemmer->stem('falar');    // fal
  • $protectedWords: words that are never stemmed, for proper nouns, acronyms and domain terms. They are compared after lowercasing, so list them in lowercase.
  • $extraExceptions: extra exceptions per suffix, keyed by the suffix of the rule, merged with the built-in ones.

Method Reference​

stem(string $word): string​

Returns the stem of a single word.