All Downloads are FREE. Search and download functionalities are using the official Maven repository.

org.cleartk.timeml.eval.Model Maven / Gradle / Ivy

There is a newer version: 3.0.0
Show newest version
/*
 * Copyright (c) 2013, Regents of the University of Colorado 
 * All rights reserved.
 * 
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions are met:
 * 
 * Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer. 
 * Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. 
 * Neither the name of the University of Colorado at Boulder nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission. 
 * 
 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
 * POSSIBILITY OF SUCH DAMAGE. 
 */
package org.cleartk.timeml.eval;

import java.io.File;
import java.util.Collection;
import java.util.List;
import java.util.Map;
import java.util.Set;

import org.apache.uima.analysis_component.AnalysisComponent;
import org.apache.uima.analysis_engine.AnalysisEngineDescription;
import org.apache.uima.cas.Feature;
import org.apache.uima.jcas.JCas;
import org.apache.uima.jcas.cas.TOP;
import org.apache.uima.jcas.tcas.Annotation;
import org.apache.uima.resource.ResourceInitializationException;
import org.cleartk.classifier.CleartkSequenceAnnotator;
import org.cleartk.classifier.DataWriter;
import org.cleartk.classifier.SequenceDataWriter;
import org.cleartk.classifier.jar.DefaultDataWriterFactory;
import org.cleartk.classifier.jar.DefaultSequenceDataWriterFactory;
import org.cleartk.classifier.jar.DirectoryDataWriterFactory;
import org.cleartk.classifier.jar.GenericJarClassifierFactory;
import org.cleartk.classifier.jar.JarClassifierBuilder;
import org.cleartk.classifier.jar.Train;
import org.cleartk.classifier.viterbi.DefaultOutcomeFeatureExtractor;
import org.cleartk.classifier.viterbi.ViterbiDataWriterFactory;
import org.cleartk.eval.AnnotationStatistics;
import org.cleartk.timeml.type.Anchor;
import org.cleartk.timeml.type.TemporalLink;
import org.uimafit.factory.AnalysisEngineFactory;
import org.uimafit.util.JCasUtil;

import com.google.common.base.Function;
import com.google.common.base.Joiner;
import com.google.common.base.Objects;
import com.google.common.base.Predicate;
import com.google.common.collect.Iterables;
import com.google.common.collect.Lists;
import com.google.common.collect.Maps;
import com.google.common.collect.Sets;

/**
 * Stores information needed for training and evaluating a machine learning model
 * 
 * 
* Copyright (c) 2013, Regents of the University of Colorado
* All rights reserved. * * @author Steven Bethard */ class Model { public static class Params { public Class dataWriterClass; public int nViterbiOutcomes; public String[] trainingArguments; public Params(Class dataWriterClass, String... trainingArguments) { this(dataWriterClass, 0, trainingArguments); } public Params(Class dataWriterClass, int nViterbiOutcomes, String ... trainingArguments) { this.dataWriterClass = dataWriterClass; this.nViterbiOutcomes = nViterbiOutcomes; this.trainingArguments = trainingArguments; } @Override public String toString() { Objects.ToStringHelper helper = Objects.toStringHelper(this.getClass()); helper.add("dataWriterClass", this.dataWriterClass.getSimpleName()); if (this.nViterbiOutcomes > 0) { helper.add("nViterbiOutcomes", this.nViterbiOutcomes); } if (this.trainingArguments.length > 0) { helper.add("trainingArguments", Joiner.on(' ').join(this.trainingArguments)); } return helper.toString(); } } public enum EvaluationType { NORMAL, GOLD_SPANS, SYSTEM_SPANS, INTERSECTED_SPANS } public enum LoggingType { NONE, SYSTEM_PREDICTIONS } public static final File DEFAULT_DIRECTORY = null; public String name; public List> prerequisites; private Class annotatorClass; public Model.Params bestParams; public List paramsToSearch; private Model.EvaluationType evaluationType; private Model.LoggingType loggingType; private Class annotationClass; Function annotationToSpan; private Function annotationToOutcome; private String featureToRemove; public Model( String name, List> prerequisites, Class annotatorClass, Model.Params bestParams, List paramsToSearch, Model.EvaluationType evaluationType, Model.LoggingType loggingType, Class annotationClass, Function annotationToSpan, Function annotationToOutcome, String featureToRemove) { this.name = name; this.prerequisites = prerequisites; this.annotatorClass = annotatorClass; this.bestParams = bestParams; this.paramsToSearch = paramsToSearch; this.evaluationType = evaluationType; this.loggingType = loggingType; this.annotationClass = annotationClass; this.annotationToSpan = annotationToSpan; this.annotationToOutcome = annotationToOutcome; this.featureToRemove = featureToRemove; } @Override public String toString() { Objects.ToStringHelper helper = Objects.toStringHelper(this.getClass()); helper.add("name", this.name); return helper.toString(); } public AnalysisEngineDescription getWriterDescription(File directory, Model.Params params) throws ResourceInitializationException { AnalysisEngineDescription desc; if (params.nViterbiOutcomes > 0) { desc = AnalysisEngineFactory.createPrimitiveDescription( this.annotatorClass, CleartkSequenceAnnotator.PARAM_DATA_WRITER_FACTORY_CLASS_NAME, ViterbiDataWriterFactory.class, ViterbiDataWriterFactory.PARAM_DELEGATED_DATA_WRITER_FACTORY_CLASS, DefaultDataWriterFactory.class, DefaultDataWriterFactory.PARAM_DATA_WRITER_CLASS_NAME, params.dataWriterClass, ViterbiDataWriterFactory.PARAM_OUTCOME_FEATURE_EXTRACTOR_NAMES, DefaultOutcomeFeatureExtractor.class, DefaultOutcomeFeatureExtractor.PARAM_MOST_RECENT_OUTCOME, 1, DefaultOutcomeFeatureExtractor.PARAM_LEAST_RECENT_OUTCOME, params.nViterbiOutcomes, DirectoryDataWriterFactory.PARAM_OUTPUT_DIRECTORY, this.getModelDirectory(directory, params)); } else { String datatWriterParamName; if (SequenceDataWriter.class.isAssignableFrom(params.dataWriterClass)) { datatWriterParamName = DefaultSequenceDataWriterFactory.PARAM_DATA_WRITER_CLASS_NAME; } else if (DataWriter.class.isAssignableFrom(params.dataWriterClass)) { datatWriterParamName = DefaultDataWriterFactory.PARAM_DATA_WRITER_CLASS_NAME; } else { throw new RuntimeException("Invalid data writer class: " + params.dataWriterClass); } desc = AnalysisEngineFactory.createPrimitiveDescription( this.annotatorClass, datatWriterParamName, params.dataWriterClass, DirectoryDataWriterFactory.PARAM_OUTPUT_DIRECTORY, this.getModelDirectory(directory, params)); } return desc; } public void train(File directory, Model.Params params) throws Exception { Train.main(this.getModelDirectory(directory, params), params.trainingArguments); } public void cleanTrainingFiles(File directory, Model.Params params) { this.cleanTrainingFiles(this.getModelDirectory(directory, params)); } private void cleanTrainingFiles(File directory) { File modelJarFile = JarClassifierBuilder.getModelJarFile(directory); for (File file : directory.listFiles()) { if (file.isDirectory()) { this.cleanTrainingFiles(file); } if (!file.equals(modelJarFile)) { file.delete(); } } } public AnalysisEngineDescription getAnnotatorDescription(File directory, Model.Params params) throws ResourceInitializationException { return AnalysisEngineFactory.createPrimitiveDescription( this.annotatorClass, GenericJarClassifierFactory.PARAM_CLASSIFIER_JAR_PATH, JarClassifierBuilder.getModelJarFile(this.getModelDirectory(directory, params))); } public void evaluate(JCas goldView, JCas systemView, AnnotationStatistics stats) { Collection goldAnnotations = this.select(goldView); Collection systemAnnotations = this.select(systemView); final Set spans; switch (this.evaluationType) { case NORMAL: spans = null; break; case GOLD_SPANS: spans = Sets.newHashSet(Iterables.transform(this.select(goldView), this.annotationToSpan)); break; case SYSTEM_SPANS: spans = Sets.newHashSet(Iterables.transform(this.select(systemView), this.annotationToSpan)); break; case INTERSECTED_SPANS: spans = Sets.intersection( Sets.newHashSet(Iterables.transform(this.select(goldView), this.annotationToSpan)), Sets.newHashSet(Iterables.transform(this.select(systemView), this.annotationToSpan))); break; default: throw new IllegalArgumentException("Unknown evaluation type: " + this.evaluationType); } if (spans != null) { Predicate isSelectedSpan = new Predicate() { @Override public boolean apply(ANNOTATION_TYPE annotation) { return spans.contains(annotationToSpan.apply(annotation)); } }; goldAnnotations = Lists.newArrayList(Iterables.filter(goldAnnotations, isSelectedSpan)); systemAnnotations = Lists.newArrayList(Iterables.filter(systemAnnotations, isSelectedSpan)); } switch (this.loggingType) { case NONE: break; case SYSTEM_PREDICTIONS: Map goldMap = Maps.newHashMap(); for (ANNOTATION_TYPE annotation : goldAnnotations) { goldMap.put(this.annotationToSpan.apply(annotation), annotation); } Map systemMap = Maps.newHashMap(); for (ANNOTATION_TYPE annotation : systemAnnotations) { systemMap.put(this.annotationToSpan.apply(annotation), annotation); } for (Object span : Sets.union(goldMap.keySet(), systemMap.keySet())) { ANNOTATION_TYPE goldAnnotation = goldMap.get(span); ANNOTATION_TYPE systemAnnotation = systemMap.get(span); String goldOutcome = goldAnnotation == null ? null : this.annotationToOutcome.apply(goldAnnotation); String systemOutcome = systemAnnotation == null ? null : this.annotationToOutcome.apply(systemAnnotation); if (goldAnnotation == null) { System.err.printf("%s: System added %s\n", this.name, format(systemAnnotation)); } else if (systemAnnotation == null) { System.err.printf("%s: System missed %s\n", this.name, format(goldAnnotation)); } else if (!goldOutcome.equals(systemOutcome)) { String message = "%s: System misclassified %s as %s\n"; System.err.printf(message, this.name, goldOutcome, format(systemAnnotation)); } else { System.err.printf("%s: System found %s\n", this.name, format(systemAnnotation)); } } } stats.add(goldAnnotations, systemAnnotations, this.annotationToSpan, this.annotationToOutcome); } private String format(ANNOTATION_TYPE annotation) { return this.format(annotation, this.annotationToOutcome.apply(annotation)); } private String format(TOP top, String outcome) { String result; String text = top.getCAS().getDocumentText().replaceAll("[\r\n]", " "); int nChars = 30; if (top instanceof TemporalLink) { TemporalLink link = (TemporalLink) top; Anchor source = link.getSource(); Anchor target = link.getTarget(); result = String.format( "%s(%s, %s)", link.getRelationType(), format(source, null), format(target, null)); } else if (top instanceof Annotation) { Annotation annotation = (Annotation) top; int begin = annotation.getBegin(); int end = annotation.getEnd(); int preBegin = Math.max(begin - nChars, 0); int postEnd = Math.min(end + nChars, text.length()); result = String.format( "...%s[%s%s]%s...", text.substring(preBegin, begin), text.substring(begin, end), outcome == null ? "" : "=" + outcome, text.substring(end, postEnd)); } else { throw new IllegalArgumentException("unsupported annotation type: " + top); } return result; } public Map removeModelAnnotations(JCas jCas) { Map annotations = Maps.newHashMap(); for (ANNOTATION_TYPE annotation : this.select(jCas)) { if (this.featureToRemove != null) { Feature feature = annotation.getType().getFeatureByBaseName(this.featureToRemove); annotations.put(annotation, annotation.getFeatureValueAsString(feature)); annotation.setFeatureValueFromString(feature, null); } else { annotations.put(annotation, null); annotation.removeFromIndexes(); } } return annotations; } public void restoreModelAnnotations(JCas jCas, Map annotations) { for (TOP annotation : annotations.keySet()) { String value = annotations.get(annotation); if (this.featureToRemove != null) { if (value != null) { Feature feature = annotation.getType().getFeatureByBaseName(this.featureToRemove); annotation.setFeatureValueFromString(feature, value); } } else { annotation.addToIndexes(); } } } private List select(JCas jCas) { // should restrict ourselves to the Text element, but restricting to exact // class should exclude DocumentCreationTimes, the one thing outside the Text element List annotations = Lists.newArrayList(); for (ANNOTATION_TYPE annotation : JCasUtil.select(jCas, this.annotationClass)) { if (annotation.getClass().equals(this.annotationClass)) { annotations.add(annotation); } } return annotations; } private File getModelDirectory(File directory, Model.Params params) { if (directory == DEFAULT_DIRECTORY) { String path = this.annotatorClass.getName().toLowerCase().replace('.', '/'); return new File("src/main/resources/" + path); } else { String dataWriterName = params.dataWriterClass.getSimpleName(); String viterbi = params.nViterbiOutcomes > 0 ? "viterbi" + params.nViterbiOutcomes + "_": ""; String fileName = viterbi + Joiner.on("_").join(params.trainingArguments); return new File(new File(new File(directory, this.name), dataWriterName), fileName); } } }