java-similarity

A Java library offering similarity algorithms.

---

Algorithms & usages

Cosine similarity

It's possible to perform a cosine similarity on two texts using the CosineSimilarity class:

import com.lorenzofritzsch.similarity.cosine.CosineSimilarity;

class Example {

    public void cosineSimilarityExample() {
        String text0 = "Some text";
        String text1 = "Some other text";
        int nGram = 3;

        double cosineSimilarity = CosineSimilarity.compute(text0, text1, nGram);
    }
}

Note that the compute method requires a nGram parameter. This parameter specifies the size of each n-gram, allowing for a more precise calculation of the similarity. For instance, the string "Some text" with nGram set to 3 will be split into the following list:

["Som", "ome", "me ", "e t", " te", "tex", "ext"]

allowing for a more granular check. However, you can set nGram to 0 and the cosine similarity will be performed word-for-word.

The CosineSimilarity class also offers a method for performing similarities on a large number of texts:

import com.lorenzofritzsch.similarity.cosine.CosineSimilarity;

import java.util.List;
import java.util.concurrent.Future;

class Example {

    public void cosineSimilarityExample() {
        String terms = "Some text";
        String texts = List.of("...");
        int nGram = 3;

        Map<String, Future<Double>> similarities = CosineSimilarity.compute(terms, texts, nGram);
    }
}

Under the hood, the CosineSimilarity#compute method computes the similarity between each text in texts and terms in a dedicated virtual thread, and stores the result in a Future, which is the value of the returned Map, while the key is the text the similarity was performed on.

Levenshtein distance

Use the LevenshteinDistance class to calculate the Levenshtein distance between two string:

import com.lorenzofritzsch.similarity.levenshtein.LevenshteinDistance;

class Example {

    public LevenshteinDistanceExample() {
        String a = "kitten";
        String b = "sitting";

        int levenshteinDistance = LevenshteinDistance.compute(a, b);
    }
}