Matching two strings

[nayan-jyoti]

Matching two strings

[manishobhatia]

Hi Nayan,

The usage of the library looks accurate .

1. Regarding 2 results being displayed. The intention is to show which documents match with others. If you prefer you could use applyMatchByGroups which will club all the matching elements in a single group and display it.
2. The numbers as input should not be a problem to match. I think your match result might have fallen below the default threshold (0.5) . Each element is separated by a space and matched with others, in the example "Nayan J Bayan 123" it will have 4 tokens , so if you have a matching element with more than 2 tokens that are similar it should match.

Here are modifications to your example with applyMatchByGroups

       String[] input = new String[]{"Nayan J 123", "Nayan J Bayan 123"};
        
        Document document = null;
        List<Document> documentList = new ArrayList<>();
        for (String str : input) {
            document = new Document.Builder(str)
                    .addElement(new Element.Builder<String>().setValue(str).setType(ElementType.TEXT).createElement())
                    .createDocument();
            documentList.add(document);
        }
        Set<Set<Match<Document>>> result = matchService.applyMatchByGroups(documentList);

        result.forEach(entry -> {
            entry.forEach(match -> {
                System.out.println("Data: " + match.getData() + " Matched With: " + match.getMatchedWith() + " Score: "
                        + match.getScore().getResult());
            });
        });

Hope this helps

Thanks

[nayan-jyoti]

Hi Manish,

Thanks for prompt response. It was indeed helpful.

document = new Document.Builder(str).addElement(new Element.Builder<String>().setValue(str) .setType(ElementType.TEXT).setThreshold(0.0).createElement()).createDocument();

I played with few more string pairs after setting threshold value to 0.0 like above:

1. "Vrij Bhooshan" & "VRIJA BHOOSHAN"
2. "Mohammad Ashfaque" & "MOHAMMED ASHFAQUE"
3. "Nayan Bayan" & "Nayan Jyoti"
4. "Nayan Jyoti Bayan Test" & "Nayan Jyoti B T"

However none of them returned any result. Is it expected behaviour?

Edit:

1. I tried to set match type to Nearest Neighbour. However on running, I got below exception:
   com.intuit.fuzzymatcher.exception.MatchException: Data Type not supported
   Tried the same with ElementType.NAME, met same exception.
2. Changed the element type to NAME and was able to get very good score of 1.0 for pairs 1 and 2.
   However no luck with the rest of two.

Thanks

[manishobhatia]

Hi Nayan,

ElementType.NAME is a better choice for this kind of match. It uses Soundex to match names, and negates any misspelled or closely spelled differences in names.
Nearest Neighbors are a better choice for numeric and date type elements where values are near each other and not the same.

I see you have reduced Threshold value. If you use that on the document instead of Element, you will see them match

String[] input = new String[]{"Nayan Jyoti Bayan Test", "Nayan Jyoti B T"};

        Document document = null;
        List<Document> documentList = new ArrayList<>();
        for (String str : input) {
            document = new Document.Builder(str)
                    .addElement(new Element.Builder<String>().setValue(str).setType(ElementType.NAME).createElement())
                    .setThreshold(0.49)
                    .createDocument();
            documentList.add(document);
        }
        Set<Set<Match<Document>>> result = matchService.applyMatchByGroups(documentList);

        result.forEach(entry -> {
            entry.forEach(match -> {
                System.out.println("Data: " + match.getData() + " Matched With: " + match.getMatchedWith() + " Score: "
                        + match.getScore().getResult());
            });
        });

[nayan-jyoti]

Thanks for the explanation. Able to get scores after setting threshould at document level.
Thank you again for your time.

Closing the issue as my queries are answered