Text portuguesesnowballstemmer stem

PortugueseSnowballStemmer.Stem

The Snowball stem of one Portuguese word.

public static string Stem(string word)

Parameters — word is a single Portuguese word. It is lowercased and NFC-normalised before the rules run.

Returns — string, the stem, always lowercase. Unlike the Spanish stemmer, this one does not strip accents: 19 of the 105 pinned words stem to something still carrying one.

Exceptions — ArgumentNullException when word is null. An empty string, or a word of one character, is returned lowercased and otherwise untouched.

Example — a derivation collapsing onto its root, and a plural that does not collapse at all.

using Lodestar.Text.Stemming;

string noun = PortugueseSnowballStemmer.Stem("atividade");  // => ativ
string adverb = PortugueseSnowballStemmer.Stem("ativamente");  // => ativ
string singular = PortugueseSnowballStemmer.Stem("geração");  // => geraçã
string plural = PortugueseSnowballStemmer.Stem("gerações");  // => geraçõ

Remarks — atividade and ativamente meeting at ativ is the behaviour that makes a stemmer worth running: a noun and an adverb built from one adjective are one term to a reader and would be two to an index without it.

geração and gerações are the counter-example, and the one to know about. They are the singular and plural of the same noun and they stem to different keys — geraçã and geraçõ — because the nasal ã/õ distinction survives the rules that would otherwise have merged them. A search for one will not find the other. amável and amáveis split the same way.

This is faithful to Snowball and to nltk, not a defect here, but it means Portuguese recall is not uniform across a vocabulary. Where it matters, folding ã/õ together before indexing is the caller's decision to make, and it changes the keys on both sides of the search.

Applies to — net10.0, netstandard2.0.

See also — PortugueseSnowballStemmer, SpanishSnowballStemmer.Stem, the stemming index.