Strings partial matching

Question

Possible Duplicate:
Similarity String Comparison in Java

Hi all

I am trying to find partial matching between two Strings using Java, of course, there are lot of questions and answers on stackoverflow, however, non of these fulfill my requirment. I have two strings (sentences), for example, "strong java programming" and "Strong programing skill". Now I want to measure the degree of similarity between these two sentences like 25% not just partial matching = ture or false.

thanks

What do you mean by a "25% match?" How are you defining the percentage of a partial match? And why are you assuming there is a Java API for it? — templatetypedef, Jul 02 '12 at 21:47
you need to provide more details. What qualifies as a match? Is it just the overlap of words? is "strong programming" and "programming strong" a 100% match? Or are you doing some sort of weight analysis based on word order? — Yevgeny Simkin, Jul 02 '12 at 21:49
do you mean the percentage of consecutive characters they have in common? — user1329572, Jul 02 '12 at 21:49
by 25% match means these two sentences might be similar in %25 of their textual structure. — solidfox, Jul 02 '12 at 21:50
user1329572 that correct and weight analysis , stemming would help. — solidfox, Jul 02 '12 at 21:52
"textual structure" I meant terms in side the sentences like 'program' 'Java' — solidfox, Jul 02 '12 at 21:56

score 6 · Answer 1 · answered Jul 02 '12 at 21:49

6

You can use string distance determination algorithms like Levenshtein distance or Jaro-Winkler.

answered Jul 02 '12 at 21:49

MRalwasser

15,605
15
101
147

tsm · Answer 2 · 2012-07-03T01:46:50.483

3

Just use the String API and your own algorithms. Something like this:

public static double similarity(String a, String b) {
  double count = 0;
  String[] words = a.split();
  for(String word : words) {
    if(b.indexOf(word) != -1) {
      count++;
    }
  }
  return count / words.length;
}

The catch is that that's not quite right--you want to do a better job of looking at the words in B. I just wanted to give you a general idea of what methods and structure you might want to have. You also want to sanitize your input--make it all lower case, remove punctuation, who knows.

edited Jul 03 '12 at 01:46

answered Jul 02 '12 at 21:49

tsm

3,598
2
21
35

2

Do you mean `words.length`, not `a.length()`? And in either case, you need to check for divide-by-zero. Otherwise, you beat me to it (+1) ;-) – DNA Jul 02 '12 at 21:52
Good catch (I was even thinking arrays (I did have `length` instead of `length()`)...just flubbed it). – tsm Jul 03 '12 at 01:48

score 1 · Answer 3 · answered Jul 02 '12 at 21:51

1

You can take a look at this library : SimMetrics .

SimMetrics is a Similarity Metric Library, e.g. from edit distance's (Levenshtein, Gotoh, Jaro etc) to other metrics, (e.g Soundex, Chapman).

answered Jul 02 '12 at 21:51

aleroot

71,077
30
176
213

You can apply these metrics algorithms in this field(to compare similarity of strings), one of the common application is data mining ... – aleroot Jul 02 '12 at 21:55

Strings partial matching

3 Answers3