How To Calculate Tf-Idf and Cosine Similarity using JAVA.

Get real time news update from your favorite websites.
Don't miss any news about your favorite topic.
Personalize your app.

Check out NTyles.

Get it on....

NOTE: Lucene 4.x users please do refer
Calculate Cosine Similarity Using Lucene

For beginners doing a project in text mining aches them a lot by various term like :

• TF-IDF
• COSINE SIMILARITY
• CLUSTERING
• DOCUMENT VECTORS

In my earlier post I showed you guys what is Cosine Similarity. I will not talk about Cosine Similarity in this post but rather I will show a nice little code to calculate Cosine Similarity in java.

Many of you must be familiar with Tf-Idf(Term frequency-Inverse Document Frequency).
I will enlighten them in brief.

Term Frequency:
Suppose for a document "Tf-Idf Brief Introduction" there are overall 60000 words and a word Term-Frequency occurs 60 times.
Then , mathematically, its Term Frequency, TF = 60/60000 =0.001.

Inverse Document Frequency:
Suppose one bought Harry-Potter series, all series. Suppose there are 7 series and a word "AbraKaDabra" comes in 2 of the series.
Then, mathematically, its Inverse-Document Frequency , IDF = 1 + log(7/2) = .......(calculated it guys, don't be lazy, I am lazy not you guys.)

And Finally, TFIDF = TF * IDF;

By mathematically I assume you now know its meaning physically.

Document Vector:
There are various ways to calculate document vectors. I am just giving you an example. Suppose If I calculate all the term's TF-IDF of a document A and store them in an array(list, matrix ... in any ordered way, .. you guys are genius you know how to create a vector. ) then I get an Document Vector of TF-IDF scores of document A.

The class shown below calculates the Term Frequency(TF) and Inverse Document Frequency(IDF).

//TfIdf.java
package com.computergodzilla.tfidf;

import java.util.List;

/**
 * Class to calculate TfIdf of term.
 * @author Mubin Shrestha
 */
public class TfIdf {
    
    /**
     * Calculates the tf of term termToCheck
     * @param totalterms : Array of all the words under processing document
     * @param termToCheck : term of which tf is to be calculated.
     * @return tf(term frequency) of term termToCheck
     */
    public double tfCalculator(String[] totalterms, String termToCheck) {
        double count = 0;  //to count the overall occurrence of the term termToCheck
        for (String s : totalterms) {
            if (s.equalsIgnoreCase(termToCheck)) {
                count++;
            }
        }
        return count / totalterms.length;
    }

    /**
     * Calculates idf of term termToCheck
     * @param allTerms : all the terms of all the documents
     * @param termToCheck
     * @return idf(inverse document frequency) score
     */
    public double idfCalculator(List allTerms, String termToCheck) {
        double count = 0;
        for (String[] ss : allTerms) {
            for (String s : ss) {
                if (s.equalsIgnoreCase(termToCheck)) {
                    count++;
                    break;
                }
            }
        }
        return 1 + Math.log(allTerms.size() / count);
    }
}

The class shown below parsed the text documents and split them into tokens. This class will communicate with TfIdf.java class to calculated TfIdf. It also calls CosineSimilarity.java class to calculated the similarity between the passed documents.

//DocumentParser.java

package com.computergodzilla.tfidf;

import java.io.BufferedReader;
import java.io.File;
import java.io.FileNotFoundException;
import java.io.FileReader;
import java.io.IOException;
import java.util.ArrayList;
import java.util.List;

/**
 * Class to read documents
 *
 * @author Mubin Shrestha
 */
public class DocumentParser {

    //This variable will hold all terms of each document in an array.
    private List termsDocsArray = new ArrayList<>();
    private List allTerms = new ArrayList<>(); //to hold all terms
    private List tfidfDocsVector = new ArrayList<>();

    /**
     * Method to read files and store in array.
     * @param filePath : source file path
     * @throws FileNotFoundException
     * @throws IOException
     */
    public void parseFiles(String filePath) throws FileNotFoundException, IOException {
        File[] allfiles = new File(filePath).listFiles();
        BufferedReader in = null;
        for (File f : allfiles) {
            if (f.getName().endsWith(".txt")) {
                in = new BufferedReader(new FileReader(f));
                StringBuilder sb = new StringBuilder();
                String s = null;
                while ((s = in.readLine()) != null) {
                    sb.append(s);
                }
                String[] tokenizedTerms = sb.toString().replaceAll("[\\W&&[^\\s]]", "").split("\\W+");   //to get individual terms
                for (String term : tokenizedTerms) {
                    if (!allTerms.contains(term)) {  //avoid duplicate entry
                        allTerms.add(term);
                    }
                }
                termsDocsArray.add(tokenizedTerms);
            }
        }

    }

    /**
     * Method to create termVector according to its tfidf score.
     */
    public void tfIdfCalculator() {
        double tf; //term frequency
        double idf; //inverse document frequency
        double tfidf; //term requency inverse document frequency        
        for (String[] docTermsArray : termsDocsArray) {
            double[] tfidfvectors = new double[allTerms.size()];
            int count = 0;
            for (String terms : allTerms) {
                tf = new TfIdf().tfCalculator(docTermsArray, terms);
                idf = new TfIdf().idfCalculator(termsDocsArray, terms);
                tfidf = tf * idf;
                tfidfvectors[count] = tfidf;
                count++;
            }
            tfidfDocsVector.add(tfidfvectors);  //storing document vectors;            
        }
    }

    /**
     * Method to calculate cosine similarity between all the documents.
     */
    public void getCosineSimilarity() {
        for (int i = 0; i < tfidfDocsVector.size(); i++) {
            for (int j = 0; j < tfidfDocsVector.size(); j++) {
                System.out.println("between " + i + " and " + j + "  =  "
                                   + new CosineSimilarity().cosineSimilarity
                                       (
                                         tfidfDocsVector.get(i), 
                                         tfidfDocsVector.get(j)
                                       )
                                  );
            }
        }
    }
}

This is the class that calculates Cosine Similarity:

//CosineSimilarity.java
/*
 * To change this template, choose Tools | Templates
 * and open the template in the editor.
 */
package com.computergodzilla.tfidf;

/**
 * Cosine similarity calculator class
 * @author Mubin Shrestha
 */
public class CosineSimilarity {

    /**
     * Method to calculate cosine similarity between two documents.
     * @param docVector1 : document vector 1 (a)
     * @param docVector2 : document vector 2 (b)
     * @return 
     */
    public double cosineSimilarity(double[] docVector1, double[] docVector2) {
        double dotProduct = 0.0;
        double magnitude1 = 0.0;
        double magnitude2 = 0.0;
        double cosineSimilarity = 0.0;

        for (int i = 0; i < docVector1.length; i++) //docVector1 and docVector2 must be of same length
        {
            dotProduct += docVector1[i] * docVector2[i];  //a.b
            magnitude1 += Math.pow(docVector1[i], 2);  //(a^2)
            magnitude2 += Math.pow(docVector2[i], 2); //(b^2)
        }

        magnitude1 = Math.sqrt(magnitude1);//sqrt(a^2)
        magnitude2 = Math.sqrt(magnitude2);//sqrt(b^2)

        if (magnitude1 != 0.0 | magnitude2 != 0.0) {
            cosineSimilarity = dotProduct / (magnitude1 * magnitude2);
        } else {
            return 0.0;
        }
        return cosineSimilarity;
    }
}

Here's the main class to run the code:

//TfIdfMain.java
package com.computergodzilla.tfidf;

import java.io.FileNotFoundException;
import java.io.IOException;

/**
 *
 * @author Mubin Shrestha
 */
public class TfIdfMain {
    
    /**
     * Main method
     * @param args
     * @throws FileNotFoundException
     * @throws IOException 
     */
    public static void main(String args[]) throws FileNotFoundException, IOException
    {
        DocumentParser dp = new DocumentParser();
        dp.parseFiles("D:\\FolderToCalculateCosineSimilarityOf"); // give the location of source file
        dp.tfIdfCalculator(); //calculates tfidf
        dp.getCosineSimilarity(); //calculates cosine similarity   
    }
}

You can also download the whole source code from here: Download.

Overall what I did is, I first calculate the TfIdf matrix of all the documents and then document vectors of each documents. Then I used those document vectors to calculate cosine similarity.

You think clarification is not enough. Hit me..
Happy Text-Mining!!

Please check out my first Android app, NTyles: