Stemmers

Stemmers reduce related word forms to a shared root. Improves recall; can reduce precision.

English stemmers

Type	Behaviour	Notes
`EnglishStemmer`	Porter-based	Default for English. `AnalyserFactory.Create("en")` uses it
`LightEnglishStemmer`	Lighter suffix stripping	Less aggressive than Porter
`KStemmer`	Lexicon-validated (Krovetz-inspired)	Needs `KStemLexicon.FromFile`
`HunspellStemmer`	Hunspell dictionary-based	Needs `.aff`/`.dic` files. Handles irregular forms

StemmedAnalyser also applies Porter stemming via PorterStemmerFilter.

Hunspell stemming

using Rowles.LeanCorpus.Analysis.Stemmers;

var dict = HunspellDictionary.FromFiles(new MMapDirectory("/dictionaries/en_US"));
var stemmer = new HunspellStemmer(dict);

var analyser = new StemmerAnalyser(
    tokeniser: new Tokeniser(),
    stemmer: stemmer);

Other languages

ArabicStemmer, ChineseStemmer, DutchStemmer, FrenchStemmer, GermanStemmer, ItalianStemmer, JapaneseStemmer, KoreanStemmer, PortugueseStemmer, RussianStemmer, SpanishStemmer.

Use with LanguageAnalyser or a custom pipeline.

Choosing

EnglishStemmer for ordinary English search.
LightEnglishStemmer when Porter is too aggressive.
KStemmer when false stems are worse than missed stems.
LanguageAnalyser for packaged stop words + stemming for a supported language.
Skip stemming when exact forms matter more than recall.

Example

var analyser = new LanguageAnalyser(
    tokeniser: new Tokeniser(),
    stopWords: StopWords.English,
    stemmer: new EnglishStemmer());