Text 0.3.2 dameraulevenshtein normalizeddistance

Lodestar.Text 0.3.2. 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.

DamerauLevenshtein.NormalizedDistance

Scales the distance into [0, 1] by dividing it by the length of the longer input.

public static double NormalizedDistance(ReadOnlySpan<char> a, ReadOnlySpan<char> b, TextElement element = TextElement.Utf16Unit)

Parameters — a and b are the two strings to compare. element says what counts as one character, and it also decides the lengths the result is divided by, so it moves the denominator as well as the distance.

Returns — double in [0, 1]: 0 when the two are equal, 1 when nothing at all can be reused. Two empty inputs give 0 rather than a division by zero.

Example — one swap over six characters.

using Lodestar.Text.Distances;

double d = DamerauLevenshtein.NormalizedDistance("MARTHA", "MARHTA");   // => 0.1666…

Remarks — this is the form to threshold on ("reject anything above 0.2") and the only form worth comparing across pairs of different lengths. The divisor is max(len(a), len(b)).

The trap is that "normalized" does not mean "interchangeable". Indel.NormalizedDistance divides by the sum of the two lengths instead, so the same pair scores differently on the two scales and a threshold tuned against one is meaningless against the other. Check the divisor before you move a threshold between measures.

Applies to — net10.0, netstandard2.0.

See also — DamerauLevenshtein.Distance, DamerauLevenshtein.NormalizedSimilarity, Levenshtein.NormalizedDistance, the Python equivalence table.