Golang : How to find out similarity between two strings with Jaro-Winkler Distance?


Tags : golang words-similarity approximate-string-matching jaro-winkler-distance

There are times when we need to detect if a paragraph or an article has a pattern of using multiple words with high similarities. This is useful in situation such as detecting a pattern to see if it is written by a same person, multiple persons or generated by computer programs. One way to do this is to use the Jaro-Winkler Distance algorithm and the code below demonstrate how Jaro-Winkler distance is implemented.

Here you go!


 package main

 import (
  "fmt"
  "strings"
 )

 func main() {
  fmt.Printf("example - existence -> %0.5f\n", JaroWinklerDistance("example", "existence"))
  fmt.Printf("feel - fill -> %0.5f\n", JaroWinklerDistance("feel", "fill"))
  fmt.Printf("octopus -  -> %0.5f\n", JaroWinklerDistance("octopus", ""))
  fmt.Printf("stick - stix -> %0.5f\n", JaroWinklerDistance("stick", "stix"))
  fmt.Printf("top - stop -> %0.5f\n", JaroWinklerDistance("top", "stop"))
  fmt.Printf("tick - lick -> %0.5f\n", JaroWinklerDistance("tick", "lick"))
  fmt.Printf("golang - Golang -> %0.5f\n", JaroWinklerDistance("golang", "golang"))
 }

 // JaroWinklerDistance - calculate and return the Jaro-Winkler distance
 func JaroWinklerDistance(s1, s2 string) float64 {

  s1Matches := make([]bool, len(s1)) // |s1|
  s2Matches := make([]bool, len(s2)) // |s2|

  var matchingCharacters = 0.0
  var transpositions = 0.0

  // sanity checks

  // return 0 if either one is empty string
  if len(s1) == 0 || len(s2) == 0 {
 return 0 // no similarity
  }

  // return 1 if both strings are empty
  if len(s1) == 0 && len(s2) == 0 {
 return 1 // exact match
  }

  if strings.EqualFold(s1, s2) { // case insensitive
 return 1 // exact match
  }

  // Two characters from s1 and s2 respectively,
  // are considered matching only if they are the same and not farther than
  // [ max(|s1|,|s2|) / 2 ] - 1
  matchDistance := len(s1)
  if len(s2) > matchDistance {
 matchDistance = len(s2)
  }
  matchDistance = matchDistance/2 - 1

  // Each character of s1 is compared with all its matching characters in s2
  for i := range s1 {
 low := i - matchDistance
 if low < 0 {
 low = 0
 }
 high := i + matchDistance + 1
 if high > len(s2) {
 high = len(s2)
 }
 for j := low; j < high; j++ {
 if s2Matches[j] {
 continue
 }
 if s1[i] != s2[j] {
 continue
 }
 s1Matches[i] = true
 s2Matches[j] = true
 matchingCharacters++
 break
 }
  }

  if matchingCharacters == 0 {
 return 0 // no similarity, exit early
  }

  // Count the transpositions.
  // The number of matching (but different sequence order) characters divided by 2 defines the number of transpositions
  k := 0
  for i := range s1 {
 if !s1Matches[i] {
 continue
 }
 for !s2Matches[k] {
 k++
 }
 if s1[i] != s2[k] {
 transpositions++ // increase transpositions
 }
 k++
  }
  transpositions /= 2

  weight := (matchingCharacters/float64(len(s1)) + matchingCharacters/float64(len(s2)) + (matchingCharacters-transpositions)/matchingCharacters) / 3

  //  the length of common prefix at the start of the string up to a maximum of four characters
  l := 0

  // is a constant scaling factor for how much the score is adjusted upwards for having common prefixes.
  //The standard value for this constant in Winkler's work is {\displaystyle p=0.1}p=0.1
  p := 0.1

  // make it easier for s1[l] == s2[l] comparison
  s1 = strings.ToLower(s1)
  s2 = strings.ToLower(s2)

  if weight > 0.7 {
 for (l < 4) && s1[l] == s2[l] {
 l++
 }

 weight = weight + float64(l)*p*(1-weight)
  }

  return weight
 }

Output:

 example - existence -> 0.58730
 feel - fill -> 0.66667
 octopus -  -> 0.00000
 stick - stix -> 0.84833
 top - stop -> 0.91667
 tick - lick -> 0.83333
 golang - Golang -> 1.00000

Happy coding!

References :

https://en.wikipedia.org/wiki/Jaro%E2%80%93Winkler_distance

https://github.com/jordanthomas/jaro-winkler/blob/master/index.js

  See also : Golang : Levenshtein distance example



Tags : golang words-similarity approximate-string-matching jaro-winkler-distance

By Adam Ng(黃武俊)

IF you gain some knowledge or the information here solved your programming problem. Please consider donating to the less fortunate or some charities that you like. Apart from donation, planting trees, volunteering or reducing your carbon footprint will be great too.


Advertisement