Merge pull request #376 from rzanoli/release

Release 1.1.2

README.md

@@ -1,9 +1,7 @@
 Excitement-Open-Platform
 ========================
 
-The EXCITEMENT Open Platform (EOP) is an open source software platform containing state-of-the-art algorithms for recognizing texual entailment relations:  
-
-_given two text fragments, one named text and the other named hypothesis, the task consists in recognizing whether the hypothesis can be inferred from the text_
+This repository contains both the code and the documentation (i.e. wiki pages) of the next Excitement Open Platform (EOP) release. EOP is an open source software platform containing state-of-the-art algorithms for recognizing texual entailment relations: _given two text fragments, one named text and the other named hypothesis, the task consists in recognizing whether the hypothesis can be inferred from the text_
 
 
 __Master Branch Status:__ [![Build Status](http://hlt-services4.fbk.eu:8080/jenkins/job/master_branch/badge/icon)](http://hlt-services4.fbk.eu:8080/jenkins/job/master_branch/)

biutee/pom.xml

@@ -4,7 +4,7 @@
   <parent>
     <groupId>eu.excitementproject</groupId>
     <artifactId>eop</artifactId>
-    <version>1.1.1</version>
+    <version>1.1.2</version>
   </parent>
   <artifactId>biutee</artifactId>
   <name>biutee</name>
@@ -36,7 +36,7 @@
 	    <dependency>
     		<groupId>eu.excitementproject</groupId>
     		<artifactId>transformations</artifactId>
-    		<version>1.1.1</version>
+    		<version>1.1.2</version>
     	</dependency>
 
   </dependencies>

biutee/src/main/java/eu/excitementproject/eop/biutee/classifiers/ClassifierFactory.java

@@ -29,19 +29,56 @@ public class ClassifierFactory
 	public static double SVM_SLACK_COEFFICIENT = BiuteeConstants.SVM_SLACK_COEFFICIENT;
 	public static double LOGISTIC_REGRESSION_LEARNING_RATE= 0.005;
 
+	public ClassifierFactory(Boolean useF1Classifier)
+	{
+		this.useF1Classifier = useF1Classifier;
+	}
+
+
+	public LinearTrainableStorableClassifier getProperClassifier(boolean datasetNatureF1) throws ClassifierException
+	{
+		return getProperClassifierForSearch(datasetNatureF1);
+	}
+
+	public LinearTrainableStorableClassifier getProperClassifierForSearch(boolean datasetNatureF1) throws ClassifierException
+	{
+		boolean f1 = false;
+		if (this.useF1Classifier!=null)
+		{
+			f1 = this.useF1Classifier.booleanValue();
+		}
+		else
+		{
+			f1 = datasetNatureF1;
+		}
+
+		if (f1)
+		{
+			logger.info("Using F1 optimized classifier");
+			return getF1Classifier();
+		}
+		else
+		{
+			logger.info("Using accuracy optimized classifier");
+			return getByPoolLinearClassifier();
+		}
+	}
+
+	@Deprecated
 	public TrainableStorableClassifier getDefaultClassifier() throws ClassifierException
 	{
 		logger.info("getDefaultClassifier()");
 		return getByPoolLinearClassifier();
 	}
 
+	@Deprecated
 	public LinearTrainableStorableClassifier getDefaultClassifierForSearch() throws ClassifierException
 	{
 		logger.info("getDefaultClassifierForSearch()");
 		return getByPoolLinearClassifier();
 	}
 
-
+	@Deprecated
 	public LinearTrainableStorableClassifier getF1Classifier() throws ClassifierException
 	{
 		logger.info("getF1Classifier() returns new GradientAscentClassifier(0.0, 0.01, new F_alpha_DerivativeCalculator(Constants.F1_CLASSIFIER_GAMMA_FOR_SIGMOID),Constants.F1_CLASSIFIER_GAMMA_FOR_SIGMOID)");
@@ -166,7 +203,8 @@ protected TrainableClassifier getDefaultSvmClassifier() throws ClassifierExcepti
 
 		return new SvmLightClassifier(trainingParameters);
 	}
+
+	private final Boolean useF1Classifier;
 
 	private static Logger logger = Logger.getLogger(ClassifierFactory.class);
-
 }

biutee/src/main/java/eu/excitementproject/eop/biutee/rteflow/endtoend/Results.java

@@ -8,6 +8,8 @@
 import eu.excitementproject.eop.biutee.classifiers.Classifier;
 import eu.excitementproject.eop.biutee.classifiers.LabeledSample;
 import eu.excitementproject.eop.biutee.utilities.BiuteeException;
+import eu.excitementproject.eop.common.datastructures.immutable.ImmutableList;
+import eu.excitementproject.eop.common.datastructures.immutable.ImmutableListWrapper;
 
 /**
  * Holds the entailment-decisions of a full dataset of T-H pairs.
@@ -96,6 +98,29 @@ public Vector<LabeledSample> getResultsAsLabeledSamples() throws BiuteeException
 		}
 		return samples;
 	}
+
+	///// Standard getters:
+
+
+	/**
+	 * Returns the proofs
+	 * @return the proofs
+	 */
+	public ImmutableList<InstanceAndProof<I, P>> getProofs()
+	{
+		return new ImmutableListWrapper<>(proofs);
+	}
+
+
+	/**
+	 * Returns the classifier for predictions
+	 * @return the classifier for predictions
+	 */
+	public Classifier getClassifierForPredictions()
+	{
+		return classifierForPredictions;
+	}
+
 
 
 	protected final List<InstanceAndProof<I, P>> proofs;

biutee/src/main/java/eu/excitementproject/eop/biutee/rteflow/endtoend/TimeStatistics.java

@@ -16,17 +16,26 @@ public class TimeStatistics implements Serializable
 	private static final long serialVersionUID = 2588315787317423702L;
 
 	public static TimeStatistics fromTimeElapsedTracker(TimeElapsedTracker tracker)
+	{
+		return fromTimeElapsedTracker(tracker,null,null);
+	}
+
+	public static TimeStatistics fromTimeElapsedTracker(TimeElapsedTracker tracker, Long numberOfExpandedElements, Long numberOfGeneratedElements)
 	{
 		long cputime = tracker.getCpuTimeElapsed();
 		long worldtime = tracker.getWorldClockElapsed();
-		return new TimeStatistics(cputime,worldtime);
+		return new TimeStatistics(cputime,worldtime,numberOfExpandedElements,numberOfGeneratedElements);
 	}
+
 
-	public TimeStatistics(long cpuTimeNanoSeconds, long worldClockTimeMilliSeconds)
+	public TimeStatistics(long cpuTimeNanoSeconds, long worldClockTimeMilliSeconds,
+			Long numberOfExpandedElements,Long numberOfGeneratedElements)
 	{
 		super();
 		this.cpuTimeNanoSeconds = cpuTimeNanoSeconds;
 		this.worldClockTimeMilliSeconds = worldClockTimeMilliSeconds;
+		this.numberOfExpandedElements = numberOfExpandedElements;
+		this.numberOfGeneratedElements = numberOfGeneratedElements;
 	}
 
 
@@ -39,18 +48,33 @@ public long getWorldClockTimeMilliSeconds()
 	{
 		return worldClockTimeMilliSeconds;
 	}
+	public Long getNumberOfExpandedElements()
+	{
+		return numberOfExpandedElements;
+	}
+	public Long getNumberOfGeneratedElements()
+	{
+		return numberOfGeneratedElements;
+	}
 
-
-
-
 
 	@Override
 	public String toString()
 	{
-		return "Cpu Time (ns) = "
+		String ret = "Cpu Time (ns) = "
 				+ String.format("%,d", getCpuTimeNanoSeconds()) 
 				+ ", World Clock Time (ms) = "
-				+ String.format("%,d",getWorldClockTimeMilliSeconds()) ;
+				+ String.format("%,d",getWorldClockTimeMilliSeconds());
+
+		if ( (numberOfExpandedElements!=null) && (numberOfGeneratedElements!=null) )
+		{
+			ret = ret
+					+ ", Number of expanded elements = "
+					+ String.format("%,d",getNumberOfExpandedElements())
+					+ ", Number of generated elements = "
+					+ String.format("%,d",getNumberOfGeneratedElements());
+		}
+		return ret;
 	}
 
 
@@ -59,4 +83,6 @@ public String toString()
 
 	private final long cpuTimeNanoSeconds;
 	private final long worldClockTimeMilliSeconds;
+	private final Long numberOfExpandedElements;
+	private final Long numberOfGeneratedElements;
 }

biutee/src/main/java/eu/excitementproject/eop/biutee/rteflow/endtoend/Trainer.java

@@ -6,14 +6,16 @@
 import static eu.excitementproject.eop.biutee.utilities.BiuteeConstants.LEARNING_MODEL_FILE_SEARCH_INDICATOR;
 
 import java.io.File;
+import java.io.FileOutputStream;
+import java.io.IOException;
+import java.io.ObjectOutputStream;
 import java.util.Iterator;
 import java.util.LinkedList;
 import java.util.List;
 import java.util.Vector;
 
 import org.apache.log4j.Logger;
 
-
 import eu.excitementproject.eop.biutee.classifiers.LabeledSample;
 import eu.excitementproject.eop.biutee.classifiers.LinearTrainableStorableClassifier;
 import eu.excitementproject.eop.biutee.classifiers.TrainableStorableClassifier;
@@ -63,29 +65,36 @@ public Trainer(Dataset<I> dataset,
 
 	public void train() throws BiuteeException
 	{
-		verifyInput();
-		classifierForSearch = classifierGenerator.createReasonableGuessClassifier();
-		iterationNumber = 0;
-		Double successRatePreviousIteration = null;
-		Double successRateCurrentIteration = null;
-		do
+		try
 		{
-			if (samplesLastIteration!=null)
+			verifyInput();
+			classifierForSearch = classifierGenerator.createReasonableGuessClassifier();
+			iterationNumber = 0;
+			Double successRatePreviousIteration = null;
+			Double successRateCurrentIteration = null;
+			do
 			{
-				samplesOfOlderIterations.add(samplesLastIteration);
+				if (samplesLastIteration!=null)
+				{
+					samplesOfOlderIterations.add(samplesLastIteration);
+				}
+				iterate();
+				storeClassifiers();
+
+				successRatePreviousIteration = successRateCurrentIteration;
+				successRateCurrentIteration = resultsLastIteration.getSuccessRate();
+				++iterationNumber;
+
+				logIterationResults();
+				endOfIterationEntryPoint();
 			}
-			iterate();
-			storeClassifiers();
-
-			successRatePreviousIteration = successRateCurrentIteration;
-			successRateCurrentIteration = resultsLastIteration.getSuccessRate();
-			++iterationNumber;
-
-			logIterationResults();
-			endOfIterationEntryPoint();
+			while (!isMainLoopDone(iterationNumber,successRatePreviousIteration,successRateCurrentIteration));
+			logger.info("Training done. Results of last iteration:\n"+resultsLastIteration.print());
+		}
+		catch (IOException e)
+		{
+			throw new BiuteeException("IO failure. Please see nested exception.",e);
 		}
-		while (!isMainLoopDone(iterationNumber,successRatePreviousIteration,successRateCurrentIteration));
-		logger.info("Training done. Results of last iteration:\n"+resultsLastIteration.print());
 	}
 
 	/**
@@ -162,7 +171,7 @@ private void iterate() throws BiuteeException
 		resultsLastIteration.compute();
 	}
 
-	private void logIterationResults() throws BiuteeException
+	private void logIterationResults() throws BiuteeException, IOException
 	{
 		logger.info("Iteration done.\nProofs:");
 		Iterator<String> detailsIterator = resultsLastIteration.instanceDetailsIterator();
@@ -172,6 +181,19 @@ private void logIterationResults() throws BiuteeException
 			logger.info(details);
 		}
 		logger.info("Current iteration result summary:\n"+resultsLastIteration.print());
+
+
+		if (BiuteeConstants.SAVE_SERIALIZED_RESULTS)
+		{
+			File serFile = new File(BiuteeConstants.RESULTS_SER_FILE_PREFIX+"_"+iterationNumber+BiuteeConstants.RESULTS_SER_FILE_POSTFIX);
+			logger.info("Iteration "+iterationNumber+" done."
+					+ " Saving results in ser file to "+serFile.getName());
+			try(ObjectOutputStream serStream = new ObjectOutputStream(new FileOutputStream(serFile)))
+			{
+				serStream.writeObject(resultsLastIteration.getProofs());
+			}
+			ExperimentManager.getInstance().register(serFile);
+		}
 	}
 
 

biutee/src/main/java/eu/excitementproject/eop/biutee/rteflow/endtoend/default_impl/AccuracyClassifierGenerator.java

@@ -17,14 +17,14 @@
  */
 public class AccuracyClassifierGenerator extends DefaultAbstractClassifierGenerator
 {
-	public AccuracyClassifierGenerator(FeatureVectorStructureOrganizer featureVectorStructure, File modelForSearch, File modelForPredictions)
+	public AccuracyClassifierGenerator(ClassifierFactory classifierFactory, FeatureVectorStructureOrganizer featureVectorStructure, File modelForSearch, File modelForPredictions)
 	{
-		super(featureVectorStructure, modelForSearch, modelForPredictions);
+		super(classifierFactory, featureVectorStructure, modelForSearch, modelForPredictions);
 	}
 
-	public AccuracyClassifierGenerator(FeatureVectorStructureOrganizer featureVectorStructure)
+	public AccuracyClassifierGenerator(ClassifierFactory classifierFactory, FeatureVectorStructureOrganizer featureVectorStructure)
 	{
-		super(featureVectorStructure);
+		super(classifierFactory, featureVectorStructure);
 	}
 
 
@@ -33,7 +33,7 @@ public LinearTrainableStorableClassifier createTrainableClassifierForSearch() th
 	{
 		try
 		{
-			return new ClassifierFactory().getDefaultClassifierForSearch();
+			return classifierFactory.getProperClassifierForSearch(false);
 		}
 		catch (ClassifierException e)
 		{
@@ -46,7 +46,7 @@ public TrainableStorableClassifier createTrainableClassifierForPredictions() thr
 	{
 		try
 		{
-			return new ClassifierFactory().getDefaultClassifier();
+			return classifierFactory.getProperClassifier(false);
 		}
 		catch (ClassifierException e)
 		{

biutee/src/main/java/eu/excitementproject/eop/biutee/rteflow/endtoend/default_impl/DefaultAbstractClassifierGenerator.java

@@ -4,6 +4,7 @@
 
 import eu.excitementproject.eop.biutee.classifiers.Classifier;
 import eu.excitementproject.eop.biutee.classifiers.ClassifierException;
+import eu.excitementproject.eop.biutee.classifiers.ClassifierFactory;
 import eu.excitementproject.eop.biutee.classifiers.LinearClassifier;
 import eu.excitementproject.eop.biutee.classifiers.LinearTrainableStorableClassifier;
 import eu.excitementproject.eop.biutee.rteflow.endtoend.ClassifierGenerator;
@@ -22,19 +23,22 @@
 public abstract class DefaultAbstractClassifierGenerator extends ClassifierGenerator
 {
 	public DefaultAbstractClassifierGenerator(
+			ClassifierFactory classifierFactory,
 			FeatureVectorStructureOrganizer featureVectorStructure,
 			File modelForSearch, File modelForPredictions)
 	{
 		super();
+		this.classifierFactory = classifierFactory;
 		this.featureVectorStructure = featureVectorStructure;
 		this.modelForSearch = modelForSearch;
 		this.modelForPredictions = modelForPredictions;
 	}
 
 
-	public DefaultAbstractClassifierGenerator(FeatureVectorStructureOrganizer featureVectorStructure)
+	public DefaultAbstractClassifierGenerator(ClassifierFactory classifierFactory, FeatureVectorStructureOrganizer featureVectorStructure)
 	{
 		super();
+		this.classifierFactory = classifierFactory;
 		this.featureVectorStructure = featureVectorStructure;
 		this.modelForSearch = null;
 		this.modelForPredictions = null;
@@ -46,7 +50,7 @@ public LinearTrainableStorableClassifier createReasonableGuessClassifier() throw
 	{
 		try
 		{
-			ReasonableGuessCreator reasonableGuessCreator = new ReasonableGuessCreator(featureVectorStructure);
+			ReasonableGuessCreator reasonableGuessCreator = new ReasonableGuessCreator(classifierFactory,featureVectorStructure);
 			reasonableGuessCreator.create();
 			return reasonableGuessCreator.getClassifier();
 		}
@@ -87,6 +91,7 @@ public Classifier loadClassifierForPredictions() throws BiuteeException
 	}
 
 
+	protected final ClassifierFactory classifierFactory;
 	protected final FeatureVectorStructureOrganizer featureVectorStructure;
 	private final File modelForSearch; // might be null
 	private final File modelForPredictions; // might be null