Lodestar.Text 0.6.0. This page is frozen at that release. Read the current documentation for what main says now. A link to a decision or a migration page follows main, and leaves the archive.

RomanianSnowballStemmer.Stem

The Snowball stem of one Romanian word.

public static string Stem(string word)

Parameters — word is a single Romanian word. It is lowercased and NFC-normalised before the rules run, so a decomposed ă and a composed one are the same letter to the algorithm. The cedilla ş and the comma-below ș are not the same letter — see the remarks below.

Returns — string, the stem, always lowercase, carrying whatever ă, â, î, ş or ș the input had.

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

Example — a verb reduced to the stem its tenses share.

using Lodestar.Text.Stemming;

string infinitive = RomanianSnowballStemmer.Stem("cânta");  // => cânt
string imperfect = RomanianSnowballStemmer.Stem("cântam");  // => cânt
string gerund = RomanianSnowballStemmer.Stem("cântând");  // => cânt

Remarks — the result most likely to look like a bug is a modern spelling that comes back almost whole. Every suffix table is written with the cedilla ş/ţ, as the published description writes them, so a word using the correct comma-below ș/ț matches only the endings that contain neither letter. informaţie reaches inform; informație reaches informaț. That is the reference's behaviour, not an oversight — and if your text is modern Romanian, replace the two letters before you call, so both spellings meet on one key.

using Lodestar.Text.Stemming;

string old = RomanianSnowballStemmer.Stem("informaţie");  // => inform
string modern = RomanianSnowballStemmer.Stem("informație");  // => informaț
string normalised = RomanianSnowballStemmer.Stem("informație".Replace('ș', 'ş').Replace('ț', 'ţ'));  // => inform

The other one worth knowing is that a stemmer produces keys, not words. muncitor, muncitori and muncitoare all reach muncit, which is also where muncit itself lands — the point of the derivational step is exactly that collision, and the algorithm has no dictionary to tell it that what remains happens to be a participle.

Applies to — net10.0, netstandard2.0.

See also — RomanianSnowballStemmer, the stemming index.