NorphoneComparator.java

// x = the sound of both sj and kj
// ^ == start of string
// $ == end of string
// [abc] == a set of characters (as in regexp)

// IMPLEMENTED
// vowels stripped, except initial vowel
// double consonants collapse into one
// ^aa -> å
// ch -> k   
// ck -> k
// [oiuaeæødy]d -> 
// dt$ -> t
// gh -> k
// gj -> j
// ^gi -> j
// hg -> k
// hj -> j
// hl -> l
// hr -> r
// kj -> x
// ki -> x
// ld -> l
// nd -> n
// ph -> f
// th -> t
// w -> v
// x -> ks
// z -> s

// NOT IMPLEMENTED
// ^c -> k 
// sj -> x
// skj -> x
// ^ei -> æ
// d -> t
// g -> k
// kei -> x
// skei -> x
// ^ky -> x
// ^sky -> x

// NOT SURE ABOUT THESE
// ^ch[aeiouy] -> x  (charlotte)
// en$ -> 

package no.priv.garshol.duke.comparators;

import no.priv.garshol.duke.Comparator;

/**
 * My own algorithm for phonetic matching of Norwegian names, inspired
 * by Metaphone.
 */
public class NorphoneComparator implements Comparator {
  
  public double compare(String s1, String s2) {
    if (s1.equals(s2))
      return 1.0;

    if (norphone(s1).equals(norphone(s2)))
      return 0.9;

    return 0.0;
  }

  public boolean isTokenized() {
    return false;
  }
  
  /**
   * Produces the Norphone key for the given string.
   */
  public static String norphone(String str) {
    if (str.length() < 1)
      return ""; // no norphone key for the empty string

    str = str.toUpperCase();
    char[] key = new char[str.length() * 2]; // could be all X-es
    int pos = 0;
    Matcher m = new Matcher(str);

    while (m.hasNext()) {
      char ch = m.next();

      // discard duplicate characters
      if (m.isNext(ch) && ch != 'A')
        ch = ' ';

      // discard vowels
      else if (isVowel(ch) && !m.atStart())
        ch = ' ';

      else {
        switch(ch) {
        case 'A': // we only come here on the first character
          if (m.isNext('A'))
            ch = '\u00C5'; // Å
          break;
          
        case 'C':
          if (m.isNext('H') || m.isNext('K')) {
            ch = 'K';
            m.skip();
          } else
            ch = 'K';
          break;

        case 'D':
          if (m.isNext('T') && m.nextIsLast()) {
            ch = 'T';
            m.skip();
          } else if (m.previousOneOf("IOUAEY\u00D8\u00C6\u00C5") && m.isLast())
            ch = ' ';
          break;
          
        case 'G':
          if (m.isNext('H'))
            m.skip(); // 'H' is silent
          else if (m.isNext('J') || (m.isNext('I') && m.atStart())) {
            ch = 'J';
            m.skip();
          }
          break;
          
        case 'H':
          if (m.isNext('J')) {
            ch = 'J';
            m.skip();
          } else if (m.isNext('L')) {
            ch = 'L';
            m.skip();
          } else if (m.isNext('G')) {
            ch = 'G';
            m.skip();
          } else if (m.isNext('R')) {
            ch = 'R';
            m.skip();
          }

          break;

        case 'K':
          if (m.isNext('J') || m.isNext('I')) {
            ch = 'X';
            m.skip();
          }
          break;

        case 'L':
          if (m.isNext('D') && m.nextIsLast()) {
            ch = 'L';
            m.skip();
          }
          break;

        case 'N':
          if (m.isNext('D'))
            m.skip();
          break;
          
        case 'P':
          if (m.isNext('H')) {
            ch = 'F';
            m.skip(); // eat the 'H'
          }
          break;  
          
        case 'T':
          if (m.isNext('H'))
            m.skip(); // 'H' is silent
          break;

        case 'W':
          ch = 'V';
          break;

        case 'X':
          key[pos++] = 'K';
          ch = 'S';
          break;
          
        case 'Z':
          ch = 'S';
          break;
        }
      }
      
      if (ch != ' ')
        key[pos++] = ch;
    }
    
    return new String(key, 0, pos);
  }

  private static boolean isVowel(char ch) {
    return (ch == 'A' || ch == 'E' || ch == 'I' || ch == 'O' || ch == 'U' ||
            ch == 'Y' || ch == '\u00C5' || ch == '\u00C6' || ch == '\u00D8');
  }
}