String similarity that rewards a shared prefix

The Winkler adjustment exists because people mistype the ends of names far more often than the beginnings, so a shared prefix is evidence. The boost is capped at four characters to keep the score bounded.

Usage

Jarow(s1, s2, p = 0.1, max_prefix = 4)

Arguments

s1, s2

Strings to compare.

p

Prefix scaling factor.

max_prefix

Longest prefix that earns a boost.

Value

List with estimate, jaro, matches, transpositions, prefix.

Details

Formula: jaro = (m/|s1| + m/|s2| + (m - t)/m)/3, jw = jaro + l p (1 - jaro).

References

Winkler, W. E. (1990). Proc Surv Res Meth Sect ASA 354-359; Jaro, M. A. (1989) JASA 84:414-420.

Examples

# Winkler's (1990) example pair: Jaro-Winkler similarity 0.961
Jarow("MARTHA", "MARHTA")$estimate
#> [1] 0.9611111