Sentiment NLP with Mallet

SentimentMallet.java

package com.aronbordin;

import ca.uwo.csd.ai.nlp.kernel.LinearKernel;
import ca.uwo.csd.ai.nlp.mallet.libsvm.SVMClassifierTrainer;
import cc.mallet.classify.*;
import cc.mallet.pipe.*;
import cc.mallet.pipe.iterator.CsvIterator;
import cc.mallet.types.Instance;
import cc.mallet.types.InstanceList;
import cc.mallet.util.Randoms;

import java.io.*;
import java.util.ArrayList;
import java.util.Iterator;
import java.util.regex.Pattern;

public class Sentiment {

    protected Pipe pipe;
    protected InstanceList trainInstances;
    protected static final String FILE_TRAIN_X = "data/train_labeled.tsv";
    protected static final String FILE_TEST = "data/test_data.tsv";
    protected static final String FILE_TRAIN_X_BIN = "data/train_labeled.bin";
    protected static final String FILE_CLASSIFIER = "data/classifier.bin";
    protected static final String FILE_TEST_OUT = "data/test.out";

    Sentiment() {
        pipe = buildPipe();
    }

    protected void importFile() {
        CsvIterator iter = null;
        try {
            iter = new CsvIterator(FILE_TRAIN_X, "\"(\\w+)\"\\s+(\\d)\\s+(.*)", 3, 2, 1);
        } catch (FileNotFoundException e) {
            e.printStackTrace();
        }

        trainInstances = new InstanceList(pipe);
        trainInstances.addThruPipe(iter);
    }

    private void saveFile() {
        trainInstances.save(new File(FILE_TRAIN_X_BIN));
    }

    protected Pipe buildPipe() {
        ArrayList<Pipe> pipeList = new ArrayList<Pipe>();

        pipeList.add(new Input2CharSequence("UTF-8"));

        Pattern patternToken = Pattern.compile("[\\p{L}\\p{N}_]+");
        pipeList.add(new CharSequence2TokenSequence(patternToken));
        pipeList.add(new TokenSequenceLowercase());
        pipeList.add(new TokenSequenceRemoveStopwords(false, false));
//        pipeList.add(new TokenSequenceRemoveNonAlpha());
        pipeList.add(new TokenSequence2FeatureSequence());

        pipeList.add(new Target2Label());
        pipeList.add(new FeatureSequence2FeatureVector());

        return new SerialPipes(pipeList);
    }

    protected void train() {
        InstanceList[] splited_data = trainInstances.split(
                new Randoms(),
                new double[]{0.7, 0.3}
        );
        ClassifierTrainer trainer;
        Classifier classifier;
        Trial trial;

/*        System.out.println("Training with MaxEntTrainer...");
        trainer = new MaxEntTrainer();
        classifier = trainer.train(splited_data[0]);
        System.out.print("Done! ");
        trial = new Trial(classifier, splited_data[1]);
        System.out.println("Accuracy: " + trial.getAccuracy());

        System.out.println("Training with NaiveBayesTrainer...");
        trainer = new NaiveBayesTrainer();
        classifier = trainer.train(splited_data[0]);
        System.out.print("Done! ");
        trial = new Trial(classifier, splited_data[1]);
        System.out.println("Accuracy: " + trial.getAccuracy());*/

        System.out.println("Training with SVMClassifierTrainer...");
        trainer = new SVMClassifierTrainer(new LinearKernel());
        classifier = trainer.train(splited_data[0]);
        System.out.print("Done! ");
        trial = new Trial(classifier, splited_data[1]);
        System.out.println("Accuracy: " + trial.getAccuracy());
    }

    protected void trainAndPredict() {
        ClassifierTrainer trainer;
        Classifier classifier = null;

        try {
            if (new File(FILE_CLASSIFIER).exists()) {
                ObjectInputStream obj = new ObjectInputStream(new FileInputStream(new File(FILE_CLASSIFIER)));
                classifier = (Classifier) obj.readObject();
                obj.close();
            } else {
//                System.out.println("Training with MaxEnt...");
//                trainer = new MaxEntTrainer();
//                classifier = trainer.train(trainInstances);

                System.out.println("Training with NaiveBayesTrainer...");
                trainer = new NaiveBayesTrainer();
                classifier = trainer.train(trainInstances);

                ObjectOutputStream obj = new ObjectOutputStream(new FileOutputStream(new File(FILE_CLASSIFIER)));
                obj.writeObject(classifier);
                obj.close();
            }
        } catch (IOException e) {
            e.printStackTrace();
        } catch (ClassNotFoundException e) {
            e.printStackTrace();
        }

        CsvIterator iterTest = null;
        try {
            iterTest = new CsvIterator(FILE_TEST, "\"(\\w+)\"\\s+(.*)", 2, 0, 1);
        } catch (FileNotFoundException e) {
            e.printStackTrace();
        }

        Iterator<Instance> testInstances = classifier.getInstancePipe().newIteratorFrom(iterTest);

        FileWriter csv = null;
        try {
            csv = new FileWriter(new File(FILE_TEST_OUT));


            csv.append("\"id\",\"sentiment\"\n");

            while (testInstances.hasNext()) {
                Instance i = testInstances.next();
                System.out.println(classifier.classify(i));

                csv.append("\"" + i.getName() + "\",");
                csv.append(classifier.classify(i).getLabelVector().getBestLabel().toString());
                csv.append("\n");
            }

            csv.flush();
            csv.close();

        } catch (IOException e) {
            e.printStackTrace();
        }


    }


    protected void readFile() {
        trainInstances = trainInstances.load(new File(FILE_TRAIN_X_BIN));
    }

    public static void main(String a[]) {
        System.out.println("Sentiment analysis\n\n\n");
        System.out.println("Reading data...");

        Sentiment snt = new Sentiment();

        if (new File(FILE_TRAIN_X_BIN).exists()) {
            snt.readFile();
        } else {
            System.out.println("\tVectoring labeled sentences...");
            snt.importFile();
            snt.saveFile();
            System.out.println("\tDone!");
        }

        snt.train();

        System.out.println("Done!");
    }
}