This is a small utility that compares two strings that it reads from stdin for similarity and emits those that exceed a given threshold. It was developed for detecting personal names that are likely to be equal.
$ echo "Donald Knuth^Donald E. Knuth" | ./strsim -t '^' -d 0.8
The invocation above would emit the line because it finds the two
strings Donald Knuth and Donald E. Knuth similar.
Several metrics for comparing strings exist where the editing distance it probably the best known. This tool implements another algorithm that is relatively robust against swapped characters and small additions and deletions. The algorithm builds for each string a set of all adjacent characters (a set of pairs) and compares these:
Let x and y be strings and xs and ys the corresponding sets of adjacent pairs from these string and ss their intersection. The similarity s of x and y is computed as
s = (2*|ss|)/(|xs|+|ys|)
where |xs| denotes the cardinality of set |xs|. Example:
x = hello
y = hallo (German for hello)
xs = {he,el,ll,lo}
ys = {ha,al,ll,lo}
ss = {ll,lo}
s = (2*2)/(4+3) = 4/7 = 0.57
./strsim -h
usage: strsim options
strsim reads lines from stdin, splits them in two halfs
and emits all lines whose half exceed a given similarity
threshold in range 0.0..1.0
options:
-t c split input lines at character c; default is tab
-d 0.8 emit lines with similarity of 0.8 or greater; default 0.9
-h emit this help to stderr
Strsim reads input line by line from stdin and splits each line into two
strings which it compares. The line is spilt at the first tab character, or
the character by option -t if provided. The threshold that needs to be
exceeded is 0.9 by default and is likewise controlled by option -d.
Strsim is implemented in Objective Caml. To build it, simply invoke Make:
$ make
The Makefile relies on ocamlbuild for the actual build process.
Christian Lindig lindig@gmail.com
This code is in the public domain.