Jump to content

Algorithm Implementation/Strings/Dice's coefficient

From Wikibooks, open books for an open world

WARNING

[edit | edit source]

Please check, if the implementation in the progamming Language you want to use meets your needs !

Some implementations (example: new unrevised javascript implementation) count equal bigrams only once, which is inadequate for string similarity (similarity of "ggggg" and "gg" will be "1" etc. ) Read the comment inside the Java Implementaion (which in my opinion seems to be correct) for this ! Could someone, who is really a specialist on this topic, review this to clarify the problem, and do this warning in a correct way ? See Discussion !

Algorithm

[edit | edit source]

Dice's coefficient measures how similar a set and another set are. It can be used to measure how similar two strings are in terms of the number of common bigrams (a bigram is a pair of adjacent letters in the string). Below we give implementations of Dice's coefficient of two strings in different programming languages.

From the Dice coefficient Wikipedia page, when taken as a string similarity measure, the coefficient may be calculated for two strings, x and y using bigrams as follows:

where nt is the number of character bigrams found in both strings, nx is the number of bigrams in string x and ny is the number of bigrams in string y.

An alternative multiset-oriented description of d is as:

2 × |AB| / (|A| + |B|)

where A, B and AB are to be understood as multisets, with A and B corresponding to the bigrams in x and y respectively, and where AB is defined as the multiset such that, if an item, x, has multiplicity a in A and b in B, then it will have multiplicity min(a,b) in AB. Also for a multiset X, the notation |X| signifies the count of items in X.

/*
 * dice coefficient = bigram overlap * 2 / (bigrams in a + bigrams in b)
 * (C) 2007 Francis Tyers 
 * Modifications made by Stefan Koshiw 2010
 * Now it outputs values [0..1]
 * Released under the terms of the GNU GPL.
 */

float dice_coefficient(wstring string1, wstring string2)
{

        set<string> string1_bigrams;
        set<string> string2_bigrams;

        //base case
        if(string1.length() == 0 || string2.length() == 0)
        {
                return 0;
        }

        for(unsigned int i = 0; i < (string1.length() - 1); i++) {      // extract character bigrams from string1
                string1_bigrams.insert(string1.substr(i, 2));
        }
        for(unsigned int i = 0; i < (string2.length() - 1); i++) {      // extract character bigrams from string2
                string2_bigrams.insert(string2.substr(i, 2));
        }

        int intersection = 0;
        
        // find the intersection between the two sets
        
        for(set<string>::iterator IT = string2_bigrams.begin(); 
            IT != string2_bigrams.end(); 
            IT++)
        {    
                intersection += string1_bigrams.count((*IT));
        }

        // calculate dice coefficient
        int total = string1_bigrams.size() + string2_bigrams.size();
        float dice = (float)(intersection * 2) / (float)total;

        return dice;
}

Haskell

[edit | edit source]

a naive, non-optimized version

import	Data.List	(intersect, nub)

diceCoefficient sa sb =
	fromIntegral (2 * length (intersect agrams bgrams)) / fromIntegral (length agrams + length bgrams)	
	where
		agrams		= bigrams sa
		bgrams		= bigrams sb
		bigrams xs	= nub $ zip xs (tail xs)
import java.util.Arrays;
import java.util.HashSet;
import java.util.Set;

//Note that this implementation is case-sensitive!
public static double diceCoefficient(String s1, String s2)
{
	Set<String> nx = new HashSet<String>();
	Set<String> ny = new HashSet<String>();

	for (int i=0; i < s1.length()-1; i++) {
		char x1 = s1.charAt(i);
		char x2 = s1.charAt(i+1);
		String tmp = "" + x1 + x2;
		nx.add(tmp);
	}
	for (int j=0; j < s2.length()-1; j++) {
		char y1 = s2.charAt(j);
		char y2 = s2.charAt(j+1);
		String tmp = "" + y1 + y2;
		ny.add(tmp);
	}

	Set<String> intersection = new HashSet<String>(nx);
	intersection.retainAll(ny);
	double totcombigrams = intersection.size();

	return (2*totcombigrams) / (nx.size()+ny.size());
}

/**
 * Here's an optimized version of the dice coefficient calculation. It takes
 * advantage of the fact that a bigram of 2 chars can be stored in 1 int, and
 * applies a matching algorithm of O(n*log(n)) instead of O(n*n).
 * 
 * <p>Note that, at the time of writing, this implementation differs from the
 * other implementations on this page. Where the other algorithms incorrectly
 * store the generated bigrams in a set (discarding duplicates), this
 * implementation actually treats multiple occurrences of a bigram as unique.
 * The correctness of this behavior is most easily seen when getting the
 * similarity between "GG" and "GGGGGGGG", which should obviously not be 1.
 * 
 * @param s The first string
 * @param t The second String
 * @return The dice coefficient between the two input strings. Returns 0 if one
 *         or both of the strings are {@code null}. Also returns 0 if one or both
 *         of the strings contain less than 2 characters and are not equal.
 * @author Jelle Fresen
 */
public static double diceCoefficientOptimized(String s, String t)
{
	// Verifying the input:
	if (s == null || t == null)
		return 0;
	// Quick check to catch identical objects:
	if (s == t)
		return 1;
        // avoid exception for single character searches
        if (s.length() < 2 || t.length() < 2)
            return 0;

	// Create the bigrams for string s:
	final int n = s.length()-1;
	final int[] sPairs = new int[n];
	for (int i = 0; i <= n; i++)
		if (i == 0)
			sPairs[i] = s.charAt(i) << 16;
		else if (i == n)
			sPairs[i-1] |= s.charAt(i);
		else
			sPairs[i] = (sPairs[i-1] |= s.charAt(i)) << 16;

	// Create the bigrams for string t:
	final int m = t.length()-1;
	final int[] tPairs = new int[m];
	for (int i = 0; i <= m; i++)
		if (i == 0)
			tPairs[i] = t.charAt(i) << 16;
		else if (i == m)
			tPairs[i-1] |= t.charAt(i);
		else
			tPairs[i] = (tPairs[i-1] |= t.charAt(i)) << 16;

	// Sort the bigram lists:
	Arrays.sort(sPairs);
	Arrays.sort(tPairs);

	// Count the matches:
	int matches = 0, i = 0, j = 0;
	while (i < n && j < m)
	{
		if (sPairs[i] == tPairs[j])
		{
			matches += 2;
			i++;
			j++;
		}
		else if (sPairs[i] < tPairs[j])
			i++;
		else
			j++;
	}
	return (double)matches/(n+m);
}

Javascript

[edit | edit source]
function getBigrams(str) {
  const bigrams = new Set();
  for (let i = 0; i < str.length - 1; i += 1) {
    bigrams.add(str.substring(i, i + 2));
  }
  return bigrams;
}

function intersect(set1, set2) {
  return new Set([...set1].filter((x) => set2.has(x)));
}

function diceCoefficient(str1, str2) {
  const bigrams1 = getBigrams(str1);
  const bigrams2 = getBigrams(str2);
  return (2 * intersect(bigrams1, bigrams2).size) / (bigrams1.size + bigrams2.size);
}
sub dice_coefficient(){
     my ($str1, $str2) = @_;
     
     return 0 if (! length $str1 || ! length $str2 );
 
     #### bigrams using REGEX.
     ##my @bigrams_str1 = $str1 =~ m/ (?= (..) ) /xmsg;
     ##my @bigrams_str2 = $str2 =~ m/ (?= (..) ) /xmsg;

     my $len1 = ( length($str1) -  1 );
     for (my $i=0; $i<$len1; $i++){
         push @bigrams_str1, substr($str1, $i, 2);
     }

     my $len2 = ( length($str2) -  1 );
     for (my $i=0; $i<$len2; $i++){
         push @bigrams_str2, substr($str2, $i, 2);
     }

     my $intersect = 0;
     STR1: for my $bigram (@bigrams_str1){
        for my $i (0 .. $#bigrams_str2){
             next unless $bigrams_str2[$i];
             if ($bigram eq $bigrams_str2[$i]){
                $intersect++;
                $bigrams_str2[$i] = undef;
                next STR1;
             }
         }
     }

     my $combinedLength = ($#bigrams_str1+1 + $#bigrams_str2+1);
     my $dice = ($intersect * 2 / $combinedLength);
   
     return $dice;
}

Python

[edit | edit source]
""" more orthodox and robust implementation """
def dice_coefficient(a, b):
    """dice coefficient 2nt/(na + nb)."""
    if not len(a) or not len(b): return 0.0
    if len(a) == 1:  a=a+u'.'
    if len(b) == 1:  b=b+u'.'
    
    a_bigram_list=[]
    for i in range(len(a)-1):
      a_bigram_list.append(a[i:i+2])
    b_bigram_list=[]
    for i in range(len(b)-1):
      b_bigram_list.append(b[i:i+2])
      
    a_bigrams = set(a_bigram_list)
    b_bigrams = set(b_bigram_list)
    overlap = len(a_bigrams & b_bigrams)
    dice_coeff = overlap * 2.0/(len(a_bigrams) + len(b_bigrams))
    return dice_coeff

""" duplicate bigrams in a word should be counted distinctly
(per discussion), otherwise 'AA' and 'AAAA' would have a 
dice coefficient of 1...
"""
def dice_coefficient(a,b):
    if not len(a) or not len(b): return 0.0
    """ quick case for true duplicates """
    if a == b: return 1.0
    """ if a != b, and a or b are single chars, then they can't possibly match """
    if len(a) == 1 or len(b) == 1: return 0.0
    
    """ use python list comprehension, preferred over list.append() """
    a_bigram_list = [a[i:i+2] for i in range(len(a)-1)]
    b_bigram_list = [b[i:i+2] for i in range(len(b)-1)]
    
    a_bigram_list.sort()
    b_bigram_list.sort()
    
    # assignments to save function calls
    lena = len(a_bigram_list)
    lenb = len(b_bigram_list)
    # initialize match counters
    matches = i = j = 0
    while (i < lena and j < lenb):
        if a_bigram_list[i] == b_bigram_list[j]:
            matches += 2
            i += 1
            j += 1
        elif a_bigram_list[i] < b_bigram_list[j]:
            i += 1
        else:
            j += 1
    
    score = float(matches)/float(lena + lenb)
    return score

Objective C

[edit | edit source]
int compare (const void * a, const void * b)
{
    return ( *(int*)a - *(int*)b );
}

+(float)diceCoefficient:(NSString*)s s2:(NSString*)t
{
    // Verifying the input:
    if (s == nil || t == nil) return 0;
    // Quick check to catch identical objects:
    if (s == t) return 1;
    // avoid exception for single character searches
    if (s.length < 2 || t.length < 2) return 0;
    
    // Create the bigrams for string s:
    int n = s.length-1;
    int sPairs[n];
    for (int i = 0; i <= n; i++)
        if (i == 0) sPairs[i] = [s characterAtIndex:i] << 16;
        else if (i == n) sPairs[i-1] |= [s characterAtIndex:i];
        else sPairs[i] = (sPairs[i-1] |= [s characterAtIndex:i]) << 16;
    
    // Create the bigrams for string t:
    int m = t.length-1;
    int tPairs[m];
    for (int i = 0; i <= m; i++)
        if (i == 0) tPairs[i] = [t characterAtIndex:i] << 16;
        else if (i == m) tPairs[i-1] |= [t characterAtIndex:i];
        else tPairs[i] = (tPairs[i-1] |= [t characterAtIndex:i]) << 16;
    
    // Sort the bigram lists:
    qsort(sPairs, n, sizeof(int), compare);
    qsort(tPairs, m, sizeof(int), compare);
    
    // Count the matches:
    int matches = 0, i = 0, j = 0;
    while (i < n && j < m) {
        if (sPairs[i] == tPairs[j]) {
            matches += 2;
            i++;
            j++;
        } else
            if (sPairs[i] < tPairs[j]) i++; else j++;
    }
    return (float)matches/(n+m);
}
# dice coefficient = bigram overlap * 2 / (bigrams in a + bigrams in b)
def dice_coefficient(a, b)
  a_bigrams = a.each_char.each_cons(2).to_a
  b_bigrams = b.each_char.each_cons(2).to_a

  overlap = (a_bigrams & b_bigrams).size

  total = a_bigrams.size + b_bigrams.size
  dice  = overlap * 2.0 / total

  dice
end
        public static double DiceCoefficient(string strA, string strB)
        {
            HashSet<string> setA = new HashSet<string>();
            HashSet<string> setB = new HashSet<string>();

            for (int i = 0; i < strA.Length - 1; ++i)
                setA.Add(strA.Substring(i,2));

            for (int i = 0; i < strB.Length - 1; ++i)
                setB.Add(strB.Substring(i,2));

            HashSet<string> intersection = new HashSet<string>(setA);
            intersection.IntersectWith(setB);

            return (2.0 * intersection.Count) / (setA.Count + setB.Count);
        }
bigrams =: (_1}.])(<@,"0)(1}.])

dice_coefficient =: (2&*@~.@%@#)@(bigrams,bigrams)
#include <string.h>

double dice_match(const char *string1, const char *string2) {

    //check fast cases
    if (((string1 != NULL) && (string1[0] == '\0')) ||
        ((string2 != NULL) && (string2[0] == '\0'))) {
        return 0;
    }
    if (string1 == string2) {
        return 1;
    }

    size_t strlen1 = strlen(string1);
    size_t strlen2 = strlen(string2);
    if (strlen1 < 2 || strlen2 < 2) {
        return 0;
    }

    size_t length1 = strlen1 - 1;
    size_t length2 = strlen2 - 1;

    double matches = 0;
    int i = 0, j = 0;

    //get bigrams and compare
    while (i < length1 && j < length2) {
        char a[3] = {string1[i], string1[i + 1], '\0'};
        char b[3] = {string2[j], string2[j + 1], '\0'};
        int cmp = strcmpi(a, b);
        if (cmp == 0) {
            matches += 2;
        }
        i++;
        j++;
    }

    return matches / (length1 + length2);
}

A possible problem with the above example is that it considers the position of the bigrams as well as their presence/absence. An alternative implementation which is independent of bigram position is given below:

#include <stdio.h>
#include <string.h>
#include <stdlib.h>

/* unsafe macro hashes bigrams into single int */
#define HASH(x,y) (((int) (x) << 16) | (y))

/* comparison function for qsort */
int
cmp ( const void *a, const void *b ) {
	int x = *(int*)a, y = *(int*)b;
	return x-y;
}

double
dice ( const char *str1, const char *str2 ) {
	int *bg1, *bg2;
	double matches;
	size_t i, strlen1, strlen2, setsize1, setsize2;

	/* make sure we've been given strings to compare and that they point to */
	/* two distinct places in memory                                        */
	if ( str1 == NULL || str2 == NULL ) { return 0; }
	if ( str1 == str2 ) { return 1; }

	/* make sure strings are long enough (must have at least two chars) */
	strlen1 = strlen(str1);
	strlen2 = strlen(str2);
	if (strlen1 < 2 || strlen2 < 2) { return 0; }

	/* allocate memory for bigram sets */	
	setsize1 = strlen1-1;
	bg1 = calloc(setsize1, sizeof(int));
	if (!bg1) { return -1; }

	setsize2 = strlen2-1;
	bg2 = calloc(setsize2, sizeof(int));
	if (!bg2) { free(bg1); return -1; }

	/* hash the strings to produce bigram sets */	
	for (i=0; i<setsize1; i++) {
		bg1[i] = HASH(str1[i], str1[i+1]); 
	}
	
	for (i=0; i<setsize2; i++) {
		bg2[i] = HASH(str2[i], str2[i+1]); 
	}

	/* sort sets for ease of comparison */
	qsort(bg1, setsize1, sizeof(int), cmp);
	qsort(bg2, setsize2, sizeof(int), cmp);

	/* compute the size of the intersection of bigram sets */
	matches = 0;
	for ( size_t i=0, j=0; i<setsize1 && j<setsize2; ) {
		if ( bg1[i] == bg2[j] ) {
			matches++;
			i++; j++;
		} else if (bg1[i] < bg2[j]) {
			i++;
		} else {
			j++;
		}
	}

	/* always remember to free your memory */
	free(bg1);
	free(bg2);

	/* compute dice */
	return (2*matches)/(setsize1+setsize2);
}