public void Cleanup()
{
// TearDown code goes here...
_comparator = null;
}
public void Init()
{
// Setup code goes here...
_comparator = new Levenshtein();
}
public double Compare(string v1, string v2)
{
if (v1.Equals(v2))
{
return(1.0);
}
if (v1.Length + v2.Length > 20 && Levenshtein.Distance(v1, v2) == 1)
{
return(0.95);
}
// tokenize
String[] t1 = StringUtils.Split(v1);
String[] t2 = StringUtils.Split(v2);
// t1 must always be the longest
if (t1.Length < t2.Length)
{
String[] tmp = t2;
t2 = t1;
t1 = tmp;
}
// penalty imposed by pre-processing
double penalty = 0;
// if the two are of unequal lengths, make some simple checks
if (t1.Length != t2.Length && t2.Length >= 2)
{
// is the first token in t1 an initial? if so, get rid of it
if ((t1[0].Length == 2 && t1[0][1] == '.') ||
t1[0].Length == 1)
{
var tmp = new String[t1.Length - 1];
for (int ix = 1; ix < t1.Length; ix++)
{
tmp[ix - 1] = t1[ix];
}
t1 = tmp;
penalty = 0.2; // impose a penalty
}
else
{
// use similarity between first and last tokens, ignoring what's
// in between
int d1 = Levenshtein.Distance(t1[0], t2[0]);
int d2 = Levenshtein.Distance(t1[t1.Length - 1], t2[t2.Length - 1]);
return((0.4 / (d1 + 1)) + (0.4 / (d2 + 1)));
}
}
// if the two are of the same length, go through and compare one by one
if (t1.Length == t2.Length)
{
// are the names just reversed?
if (t1.Length == 2 && t1[0].Equals(t2[1]) && t1[1].Equals(t2[0]))
{
return(0.9);
}
// normal one-by-one comparison
double points = 1.0 - penalty;
for (int ix = 0; ix < t1.Length && points > 0; ix++)
{
int d = Levenshtein.Distance(t1[ix], t2[ix]);
if (ix == 0 && d > 0 &&
(t1[ix].StartsWith(t2[ix]) || t2[ix].StartsWith(t1[ix])))
{
// are we at the first name, and one is a prefix of the other?
// if so, we treat this as edit distance 1
d = 1;
}
else if (d > 1 && (ix + 1) <= t1.Length)
{
// is it an initial? ie: "marius" ~ "m." or "marius" ~ "m"
String s1 = t1[ix];
String s2 = t2[ix];
// ensure s1 is the longer
if (s1.Length < s2.Length)
{
String tmp = s1;
s1 = s2;
s2 = tmp;
}
if ((s2.Length == 2 && s2[1] == '.') ||
s2.Length == 1)
{
// so, s2 is an initial
if (s2[0] == s1[0])
{
d = 1; // initial matches token in other name
}
}
}
else if (t1[ix].Length + t2[ix].Length <= 4)
{
// it's not an initial, so if the strings are 4 characters
// or less, we quadruple the edit dist
d = d * 4;
}
else if (t1[ix].Length + t2[ix].Length <= 6)
{
// it's not an initial, so if the strings are 3 characters
// or less, we triple the edit dist
d = d * 3;
}
else if (t1[ix].Length + t2[ix].Length <= 8)
{
// it's not an initial, so if the strings are 4 characters
// or less, we double the edit dist
d = d * 2;
}
points -= d * 0.1;
}
// if both are just one token, be strict
if (t1.Length == 1 && points < 0.8)
{
return(0.0);
}
return(Math.Max(points, 0.0));
}
return(0.0);
}