Text tversky

Tversky

Shared q-grams against the two sides' surpluses, weighted separately: |A∩B| / (|A∩B| + α·|A\B| + β·|B\A|).

public static class Tversky

Example — the default weights, which are Jaccard.

using Lodestar.Text.Similarity;

double symmetric = Tversky.Similarity("apple", "pineapple");  // => 0.5555…

Remarks — the general form the rest of this namespace are cases of. α = β = 1 is Jaccard, α = β = 0.5 is SorensenDice, and any α = β gives a symmetric measure that ranks the same way both do.

What the others cannot express is α ≠ β. The two surpluses are charged separately, so the measure acquires a direction: α = 1, β = 0 charges only for what the first input holds alone and therefore asks "is a contained in b", while α = 0, β = 1 asks the reverse. That is the reason to reach for this type rather than one of the four simpler ones.

The price of that generality is that the weights are the caller's, and two of the guarantees the other four offer stop holding — the result need not lie in [0, 1], and an empty input need not score 0. Tversky.Similarity says when.

Applies to — net10.0, netstandard2.0.

See also — Jaccard, Overlap, the set-similarity index.

Members

Member	What it does
Tversky.Similarity	Shared grams against each side's surplus, weighted by alpha and beta.