Development build. This page describes main, not a released package. The latest published Lodestar.Fuzzy is 0.4.0 — read its documentation.

Home › Fuzzy › Fuzzy matching

Fuzz.TokenSetRatio

The words as sets, so extra words on one side stop counting against it.

public static double TokenSetRatio(string a, string b)

public static double TokenSetRatio(string a, string b, TextElement element)

The second overload compares over element. At TextElement.CodePoint a character outside the Basic Multilingual Plane counts once, tokens split on rapidfuzz's own whitespace and sort by code point, which is rapidfuzz's score on any string; TextElement.Utf16Unit is the first overload.

Parameters — a and b are the strings to compare. element is the unit compared, in the second overload only.

Returns — double in [0, 100], computed over the intersection and the two differences of the word sets.

Exceptions — ArgumentOutOfRangeException when element is not a declared value. ArgumentException when the two strings hold more than 63,455 distinct code points above U+0020, which is what a char can rank one unit per code point; Fuzz.Ratio answers such a pair, these scorers do not yet.

Example — one side carrying words the other does not.

using Lodestar.Fuzzy;

string query = "mariners vs angels";
string candidate = "los angeles angels vs seattle mariners";

double subset = Fuzz.TokenSetRatio(query, candidate);  // => 100

Remarks — 100, because every word of the shorter side appears in the longer one. That is the most forgiving of the seven and the easiest to misuse: it will score 100 for a query that is a subset of a candidate, however much else that candidate says.

Right for "does this short label refer to this long one" — a team name against a full fixture, a brand against a product title. Wrong for deduplication, where two records differing by several words are usually two things.

Duplicated words do not help: a set counts a word once, so "the the cat" and "the cat" compare as equal sets.

A side with no words — empty, or whitespace alone — scores 0 against anything, the other side included when it has none either, as in rapidfuzz: an empty set is not a subset match.

An emoji is two UTF-16 units, so the first overload scores two different emoji that share a high surrogate as half alike, where rapidfuzz scores them 0; pass TextElement.CodePoint when the text can leave the BMP, as decision 0001 offers on every algorithm it affects.

Applies to — net10.0, netstandard2.0.

See also — Fuzz.TokenSortRatio, Fuzz.PartialTokenSetRatio.