Java tutorial
/** * This file is part of TagLine. * * TagLine is free software: you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by * the Free Software Foundation, either version 3 of the License, or * (at your option) any later version. * * TagLine is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU General Public License for more details. * * You should have received a copy of the GNU General Public License * along with TagLine. If not, see <http://www.gnu.org/licenses/>. */ package gov.va.chir.tagline; import gov.va.chir.tagline.beans.ClassifierType; import gov.va.chir.tagline.beans.Configuration; import gov.va.chir.tagline.beans.Document; import gov.va.chir.tagline.dao.AnnotationsSaver; import gov.va.chir.tagline.dao.DatasetFeatureSaver; import gov.va.chir.tagline.dao.DatasetScoredSaver; import gov.va.chir.tagline.dao.FileDao; import gov.va.chir.tagline.structure.CheckBoxIdentifier; import gov.va.chir.tagline.structure.FreeTextIdentifier; import gov.va.chir.tagline.structure.Identifier; import gov.va.chir.tagline.structure.ListIdentifier; import gov.va.chir.tagline.structure.MedsIdentifier; import gov.va.chir.tagline.structure.QuestionIdentifier; import gov.va.chir.tagline.structure.SlotFillerIdentifier; import gov.va.chir.tagline.structure.TableIdentifier; import gov.va.chir.tagline.structure.UserDefinedIdentifier; import gov.va.chir.tagline.structure.VitalsIdentifier; import java.io.File; import java.io.IOException; import java.util.Collection; import java.util.Date; import java.util.HashSet; import java.util.Set; import org.apache.commons.io.FileUtils; public class TagLineApp { private static final boolean TO_CONSOLE = true; public static void main(String[] args) { String configFilename = "config.properties"; if (args != null && args.length == 1) { configFilename = args[0]; } final Configuration config = new Configuration(configFilename); final TagLineApp tla = new TagLineApp(config); try { tla.execute(); } catch (Exception e) { e.printStackTrace(); } } private static void writeLog(final File file, final String s) throws IOException { final String output = String.format("%s%s", s, System.getProperty("line.separator")); FileUtils.writeStringToFile(file, output, true); if (TO_CONSOLE) { System.out.print(output); } } private Configuration config; public TagLineApp(final Configuration config) { this.config = config; } private void createSyntheticDataset(final File log, final File dataset, final int numDocs, final int maxLinesPerDoc, final int maxLineLength, final int numClasses, final long randomSeed) throws IOException { writeLog(log, "Creating synthetic dataset"); if (dataset == null) { throw new IllegalArgumentException("Dataset file must not be null"); } if (numDocs <= 0 || maxLinesPerDoc <= 0 || maxLineLength <= 0) { throw new IllegalArgumentException("Number of docs, maximum lines " + "per doc, and maximum line length must all be positive " + "integers greater than zero"); } // numClasses == 0 is a special case if (numClasses != 0 && numClasses < 2) { throw new IllegalArgumentException("There must be at least two classes"); } FileDao.createSyntheticTrainingDataset(dataset, numDocs, maxLinesPerDoc, maxLineLength, numClasses, randomSeed); writeLog(log, "Finished"); } private void evaluatePerformance(final File log, final File dataset, final File performance, final ClassifierType classifierType, final String... options) throws Exception { writeLog(log, "Evaluating performance"); if (dataset == null || performance == null) { throw new IllegalArgumentException("Input dataset and output performance files must not be null"); } writeLog(log, "Loading input data"); Collection<Document> docs = FileDao.loadLabeledLines(dataset, true); writeLog(log, String.format("Loaded %,d docs", docs.size())); writeLog(log, "Training and evaluating models"); final TagLineEvaluator tle = new TagLineEvaluator(docs); tle.evaluate(classifierType, options); writeLog(log, "Saving performance"); FileDao.savePerformance(performance, tle.getEvaluationSummary()); writeLog(log, "Finished"); } public void execute() throws Exception { final File log = config.getOutLog(); log.delete(); writeLog(log, "Configuration file contents:"); writeLog(log, config.getConfigContents()); writeLog(log, "---------------------------------------------------------------"); writeLog(log, ""); writeLog(log, String.format("Starting: %s", new Date().toString())); writeLog(log, ""); final TaskType taskType = config.getTaskType(); if (taskType == TaskType.CREATE) { createSyntheticDataset(log, config.getCreateFileOutputDataset(), config.getCreateNumDocs(), config.getCreateMaxLinesPerDoc(), config.getCreateMaxLineLength(), config.getCreateNumClasses(), config.getCreateRandomSeed()); } else if (taskType == TaskType.EVALUATE) { evaluatePerformance(log, config.getEvalFileInputDataset(), config.getEvalFileOutputPerformance(), config.getEvalClassifierType(), config.getEvalClassifierOptions()); } else if (taskType == TaskType.SCORE) { scoreData(log, config.getScoreFileInputDataset(), config.getScoreFileInputModel(), config.getScoreFileOutputDataset(), config.getScoreFileOutputDatasetFeatures()); } else if (taskType == TaskType.TRAIN) { trainClassifier(log, config.getTrainFileInputDataset(), config.getTrainFileOutputModel(), config.getTrainFileOutputDataset(), config.getTrainClassifierType(), config.getTrainClassifierOptions()); } else if (taskType == TaskType.IDENTIFY) { identifyStructures(log, config.getIdentifyFileInputDataset(), config.getIdentifyFileOutputAnnotations(), config.isIdentifyCheckboxes(), config.isIdentifyFreeText(), config.isIdentifyLists(), config.isIdentifyMedications(), config.isIdentifyQuestions(), config.isIdentifySlotFillers(), config.isIdentifyTables(), config.isIdentifyVitals(), config.getIdentifyStructuresUserDefined()); } else { throw new IllegalArgumentException(String.format("Unsupported task type: %s", taskType)); } writeLog(log, ""); writeLog(log, String.format("Finishing: %s", new Date().toString())); } private void identifyStructures(final File log, final File dataset, final File outputAnnotations, final boolean identifyCheckboxes, final boolean identifyFreeText, final boolean identifyLists, final boolean identifyMedications, final boolean identifyQuestions, final boolean identifySlotFillers, final boolean identifyTables, final boolean identifyVitals, final String... identifyUserDefined) throws Exception { writeLog(log, "Identifying structures"); if (dataset == null || outputAnnotations == null) { throw new IllegalArgumentException("Input dataset and output annotations must not be null"); } writeLog(log, "Loading input data"); Collection<Document> docs = FileDao.loadLabeledLines(dataset, true); writeLog(log, String.format("\tLoaded %,d docs", docs.size())); if (docs == null || docs.isEmpty()) { throw new IllegalStateException("Must have at least one document to score"); } else { writeLog(log, "Identifying structures"); final Set<Identifier> identifiers = new HashSet<Identifier>(); if (identifyCheckboxes) { identifiers.add(new CheckBoxIdentifier()); } if (identifyFreeText) { identifiers.add(new FreeTextIdentifier()); } if (identifyLists) { identifiers.add(new ListIdentifier()); } if (identifyMedications) { identifiers.add(new MedsIdentifier()); } if (identifyQuestions) { identifiers.add(new QuestionIdentifier()); } if (identifySlotFillers) { identifiers.add(new SlotFillerIdentifier()); } if (identifyTables) { identifiers.add(new TableIdentifier()); } if (identifyVitals) { identifiers.add(new VitalsIdentifier()); } if (identifyUserDefined != null) { for (String iud : identifyUserDefined) { identifiers.add(new UserDefinedIdentifier(iud)); } } final AnnotationsSaver as = new AnnotationsSaver(outputAnnotations); for (Document doc : docs) { for (Identifier identifier : identifiers) { identifier.processDocument(doc); } as.saveRecord(doc); } } writeLog(log, "Finished"); } private void scoreData(final File log, final File dataset, final File modelFile, final File outputDataset, final File outputFeatures) throws Exception { writeLog(log, "Scoring data"); if (dataset == null || modelFile == null || outputDataset == null) { throw new IllegalArgumentException("Input and Output datasets, and model files must not be null"); } writeLog(log, "Loading input data"); Collection<Document> docs = null; if (dataset.isDirectory()) { docs = FileDao.loadScoringDocs(dataset); } else { docs = FileDao.loadScoringLines(dataset, true); } writeLog(log, String.format("\tLoaded %,d docs", docs.size())); if (docs == null || docs.isEmpty()) { throw new IllegalStateException("Must have at least one document to score"); } else { writeLog(log, "Loading model"); final TagLineModel model = FileDao.loadTagLineModel(modelFile); writeLog(log, String.format("\tLoaded %,d features", model.getFeatures().size())); writeLog(log, "Loading classifier"); final TagLineScorer tls = new TagLineScorer(model); writeLog(log, "Scoring documents"); final DatasetScoredSaver dss = new DatasetScoredSaver(outputDataset); DatasetFeatureSaver dfs = null; if (outputFeatures != null) { dfs = new DatasetFeatureSaver(outputFeatures); } for (Document doc : docs) { tls.applyModel(doc); dss.saveRecord(doc); if (dfs != null) { dfs.saveRecord(doc); } } } writeLog(log, "Finished"); } private void trainClassifier(final File log, final File dataset, final File modelFile, final File outputDataset, final ClassifierType classifierType, final String... options) throws Exception { writeLog(log, "Training a classifier"); if (dataset == null || modelFile == null) { throw new IllegalArgumentException("Dataset and model files must not be null"); } writeLog(log, "Loading input data"); Collection<Document> docs = FileDao.loadLabeledLines(dataset, true); writeLog(log, String.format("Loaded %,d docs", docs.size())); if (docs == null || docs.isEmpty()) { throw new IllegalStateException("Must have at least one document to score"); } else { writeLog(log, "Loading trainer"); final TagLineTrainer tlt = new TagLineTrainer(classifierType, options); tlt.train(docs); writeLog(log, "Finished training"); // Save model writeLog(log, "Saving model"); FileDao.saveTagLineModel(modelFile, tlt.getTagLineModel()); // Save dataset (optional) if (outputDataset != null) { writeLog(log, "Saving dataset with features"); final DatasetFeatureSaver dfs = new DatasetFeatureSaver(outputDataset); dfs.saveRecords(docs); } } writeLog(log, "Finished"); } }