String distances: string_distance

Syntax

string_distance ( method, string1, string2 )

Input parameters

method	the distance method to apply (levenshtein, damerau_levenshtein, hamming, jaro_winkler)
string1	the first string operand
string2	the second string operand

Examples of valid syntaxes

string_distance(levenshtein, "foo", "bar")
string_distance(levenshtein, DS_1, DS_2)
string_distance(damerau_levenshtein, "foo", "bar")
string_distance(hamming, "foo", "bar")
string_distance(jaro_winkler, "FOO", "BAR")

Semantics for scalar operations

Returns the distance between two strings using the specified distance method.

All distance methods are symmetric (commutative), meaning that string_distance(method, string1, string2) equals string_distance(method, string2, string1).

levenshtein: Returns the Levenshtein distance, which is the minimum number of single-character edits (insertions, deletions, or substitutions) required to change one string into another.
damerau_levenshtein: Returns the Damerau–Levenshtein distance, which extends Levenshtein distance by including transpositions of adjacent characters as a single operation.
hamming: Returns the Hamming distance, which is the number of positions at which the corresponding characters are different. Both strings must have the same length.
jaro_winkler: Returns the Jaro–Winkler distance, which is a string metric measuring an edit distance between two sequences. It is a variant of the Jaro distance metric and mainly used in the area of record linkage.

For example:

string_distance(levenshtein, "foo", "fo") gives 1
string_distance(damerau_levenshtein, "bar", "baz") gives 1
string_distance(hamming, "foo", "fob") gives 1
string_distance(jaro_winkler, "FOO", "BAR") gives a value between 0 and 1

Input parameters type

method

component<string>
| string

string1, string2

dataset { measure<string> _+ }
| component<string>
| string

Result type

result

dataset { measure<number> _+ }
| component<number>
| number

Additional Constraints

The method parameter must be one of: levenshtein, damerau_levenshtein, hamming, or jaro_winkler.
For Hamming distance, both strings must have the same length.
For operations at Data Set level, the input Data Sets must have exactly one string type Measure each.

Behaviour

The operator has the behaviour of the “Operators applicable on two Scalar Values or Data Sets or Data Set Components” (see the section “Typical behaviours of the ML Operators”). If string1 and string2 are Data Sets then string_distance returns a dataset with a single measure of type number.

Examples

Given the operand datasets DS_1 and DS_2:

Input DS_1 (see structure)

Id_1	Id_2	Me_1
1	A	foo
2	B	bar

Input DS_2 (see structure)

Id_1	Id_2	Me_1
1	A	fo
2	B	baz

Example 1

DS_r := string_distance(levenshtein, DS_1, DS_2);

results in (see structure):

DS_r
Id_1	Id_2	Me_1
1	A	1
2	B	1

Example 2

DS_r := string_distance(damerau_levenshtein, DS_1, DS_2);

results in (see structure):

DS_r
Id_1	Id_2	Me_1
1	A	1
2	B	1

Example 3

DS_r := DS_1[calc Me_2 := string_distance(hamming, Me_1, "fob")];

results in (see structure):

DS_r
Id_1	Id_2	Me_1	Me_2
1	A	foo	1
2	B	bar	3