Question

I am having some problem with Serialization

Here goes my code for writing mClassifier object to a file:

FileOutputStream fileOut = new FileOutputStream("C:\\polarity.model");
ObjectOutputStream objOut = new ObjectOutputStream(fileOut);
mClassifier.compileTo(objOut);
objOut.close();

It works fine and writes stuff to the file.

But there is a catch: myClassifier object is of type DynamicLMClassifier. compileTo method above however returns an instance of LMClassifier (superclass)

Here goes my code for reading the object:

FileInputStream in = new FileInputStream("C:\\polarity.model");
ObjectInputStream ois = new ObjectInputStream(in);
mClassifier = (DynamicLMClassifier)(ois.readObject());
ois.close();

When I read the object I typecast it to DynamicLMClassifier and it too works fine but I dont get the output I desired. While reading the object again should it not be typecasted to LMClassifier rather than DynamicLMClassifier. However if I do that, compiler complains that it should be of type DynamicLMClassifier.

Can the above be problems or am I doing something wrong some where else. I mean the code without serialization is working perfectly fine and I get the desired output, I mean when the object is in memory.

EDIT: Here is the complete code (just remove the serialization part in the train() and getSentiments() method and it works as intended), Also note that in (1) With Serialization I am not calling getSentiments() and I am just training i.e. calling the train () method (2) Now i have a serialized model after (1) and I am not calling the train() method just calling getSentiment() by just commenting out appropriate code in main:

public class PolarityBasic{

    File mPolarityDir;
    String[] mCategories;
    DynamicLMClassifier<NGramProcessLM> mClassifier,readClassifier;

    PolarityBasic(String[] args) {
        System.out.println("\nBASIC POLARITY DEMO");
        mPolarityDir = new File("C:\\review_polarity","txt_sentoken");
        System.out.println("\nData Directory=" + mPolarityDir);
        mCategories = mPolarityDir.list();
        int nGram = 8;
        mClassifier 
            = DynamicLMClassifier
            .createNGramProcess(mCategories,nGram);
   }

    void run() throws ClassNotFoundException, IOException {
        train();
   }

    boolean isTrainingFile(File file) {
        return file.getName().charAt(2) != '9';  // test on fold 9
    }

    void train() throws IOException {
        int numTrainingCases = 0;
        int numTrainingChars = 0;
        System.out.println("\nTraining.");
        for (int i = 0; i < mCategories.length; ++i) {
            String category = mCategories[i];
            Classification classification
                = new Classification(category);
            File file = new File(mPolarityDir,mCategories[i]);
            File[] trainFiles = file.listFiles();
            for (int j = 0; j < trainFiles.length; ++j) {
                File trainFile = trainFiles[j];
                if (isTrainingFile(trainFile)) {
                    ++numTrainingCases;
                    String review = Files.readFromFile(trainFile,"ISO-8859-1");
                    numTrainingChars += review.length();
                    Classified<CharSequence> classified
                        = new Classified<CharSequence>(review,classification);
                    mClassifier.handle(classified);

                }
            }
        }
        FileOutputStream fileOut = new FileOutputStream("C:\\review_polarity/polarity.model");
        ObjectOutputStream objOut = new ObjectOutputStream(fileOut);
        mClassifier.compileTo(objOut);
        objOut.close();
        System.out.println("  # Training Cases=" + numTrainingCases);
        System.out.println("  # Training Chars=" + numTrainingChars);
    }


    String getSentiment(String text) {
        try{
            FileInputStream in = new FileInputStream("C:\\review_polarity/polarity.model");
            ObjectInputStream ois = new ObjectInputStream(in);
            mClassifier = (DynamicLMClassifier)(ois.readObject());
            ois.close();
        }
        catch(Exception e){}
        Classification classification = null;
        classification = readClassifier.classify(text);
        System.out.println("classification:  " + classification);
        return (classification.bestCategory());
    }

    public static void main(String[] args) {
        try {
            PolarityBasic pB = new PolarityBasic(args);
            pB.run();
            String text = null;
            text = "It was awesome !";
           System.out.println("The text \"" + text + "\" is "
         + pB.getSentiment(text));
        } catch (Throwable t) {
            System.out.println("Thrown: " + t);
            t.printStackTrace(System.out);
        }
    }

}

Answer 1

Solved the problem. Here goes the code for reading the object back (instead of one given in the question):

LMClassifier readClassifier;
FileInputStream in = new FileInputStream("C:\\polarity.model");
ObjectInputStream ois = new ObjectInputStream(in);
readClassifier = (LMClassifier)(ois.readObject());
ois.close();