Question

I’m trying to use the wordalignment in the BerkeleyAligner.jar file from http://code.google.com/p/berkeleyaligner/ in my own java class.
I have already added the .jar file into my buildpath.

What parameters does the edu.berkeley.nlp.wordAlignment.combine.CombinedAligner take?
What does the edu.berkeley.nlp.wordAlignment.combine.CombinedAligneroutput?

What i have are 2 input files that are already sentence aligned; i.e. the sentence from line number X from the sourceFile is the same (but in a different language) as the sentence from line number X of the targetFile.

import edu.berkeley.*;
import edu.berkeley.nlp.wa.mt.Alignment;
import edu.berkeley.nlp.wa.mt.SentencePair;
public class TestAlign {

BufferedReader brSrc = new BufferedReader(new FileReader ("sourceFile"));
BufferedReader brTrg = new BufferedReader(new FileReader ("targetFile"));

String currentSrcLine;

while ((currentSrcLine = brSrc.readLine()) !=null) {
    String currentTrgLine = brTrg.readline();
    // Reads into BerkeleyAligner SentencePair format.
    SentencePair src2trg = new SentencePair(sentCounter, params.get("source"),
        Arrays.asList(srcLine.split(" ")), Arrays.asList(trgLine.split(" ")));
    // How do i call the BerkeleyAligner??
    // -What parameters does the CombinedAligner takes?
    // -What does the function/class returns?
    // I assume it returns a list of strings. 
    // Is there a class in BerkeleyAligner to read the output? 
    // Please provide some example, thank you!!
    Alignment output = edu.berkeley.nlp.wordAlignment.combine.CombinedAligner
        .something.something(currentSrcLine, currentTrgLine);
    }
}

e.g. sourceFile:

this is the first line in the textfile.
that is the second line.
foo bar likes to eat bar foo.

e.g. targetFile:

Dies ist die erste Textzeile in der Datei.
das ist die zweite Zeile.
foo bar gerne bar foo essen.

Answer 1

Actual Answer

You just wanted to align text (from a target file and a source file), right?

If so, after creating a sentence pair, you did not even need to put them in a CombinedAligner.

You could get an Alignment: (SentencePair, boolean) from that. The boolean is if you want a tree alignment.

Putting it into the constructor will generate an Alignment automatically!
So simple!

This is where I got the code: http://code.google.com/p/berkeleyaligner/source/browse/trunk/src/edu/berkeley/nlp/wa/mt/Alignment.java

---

UPDATE

Unfortunately, I misunderstood your question, and posted an irrelevant response.

However, I downloaded the jar file, found CombinedAligner.class, and decompiled it.

Here’s what I got:

package edu.berkeley.nlp.wordAlignment.combine;

import edu.berkeley.nlp.mt.Alignment;
import edu.berkeley.nlp.mt.SentencePair;
import edu.berkeley.nlp.wordAlignment.PosteriorAligner;
import edu.berkeley.nlp.wordAlignment.WordAligner;
import fig.basic.Fmt;
import fig.basic.ListUtils;
import java.util.ArrayList;
import java.util.List;

public abstract  class CombinedAligner extends PosteriorAligner {

    private static final long serialVersionUID = 1;
    WordAligner wa1;
    WordAligner wa2;

    public CombinedAligner (WordAligner, WordAligner)
    public String getName()
    public Alignment alignSentencePair(SentencePair)
    public List alignSentencePairReturnAll(SentencePair)
    public void setThreshold(int)
    abstract Alignment combineAlignments(Alignment, Alignment, SentencePair)

}

It seems that the Alignment class you’re using is edu.berkeley.nlp.mt.Alignment.

Anyway, CombinedAligner is abstract, so you can’t instantiate it. And I don’t know what the .something‘s are, because there is no static method or field.

I think that what you want, however, is alignSentencePair(SentencePair).

To get this, you need to use a subclass of CombinedAligner, as CombinedAligner is abstract.

So, after poking around the files, I found these subclasses:

edu.berkeley.nlp.wordAlignment.combine.HardUnion
edu.berkeley.nlp.wordAlignment.combine.HardIntersect
edu.berkeley.nlp.wordAlignment.combine.SoftUnion
edu.berkeley.nlp.wordAlignment.combine.SoftIntersect

You can use these instead of CombinedAligner and insert your two sentences as a SentencePair!

---

After checking, I realized that WordAligner is also abstract!

package edu.berkeley.nlp.wordAlignment;

import edu.berkeley.nlp.mt.Alignment;
import edu.berkeley.nlp.mt.SentencePair;
import fig.basic.LogInfo;
import java.io.Serializable;
import java.util.ArrayList;
import java.util.HashMap;
import java.util.Iterator;
import java.util.List;
import java.util.Map;

public abstract class WordAligner implements Serializable {

private static final long serialVersionUID = 1;
protected String modelPrefix;

public WordAligner ()
public abstract String getName()
public void setThreshold(double)
public Alignment alignSentencePair(SentencePair)
public Map alignSentencePairs(List)
public Alignment thresholdAlignment(Alignment, double)
public String getModelPrefix()
public String toString()

}

I found a subclass, though:

edu.berkeley.nlp.wordAlignment.IterWordAligner

Unfortunately, this is still abstract.

But there’s a subclass of IterWordAligner that isn’t:
edu.berkeley.nlp.wordAlignment.EMWordAligner

However, the constructor is really weird.

public EMWordAligner (SentencePairState$Factory, Evaluator, boolean)

It uses an INNER CLASS in the CONSTRUCTOR!? That’s terrible programming practice.

WAIT…

I found a word aligner!
http://code.google.com/p/tdx-nlp/source/browse/trunk/pa2/java/src/cs224n/assignments/WordAlignmentTester.java?r=67

Maybe that helps and you can resolve your problem with it.