How to use

Train

in

org.cleartk.ml.jar

public static void main(String... args) throws Exception {
 String programName = Train.class.getName();
 String usage = String.format("usage: java %s DIR\n\n"
   + "The directory DIR should contain the training-data.xxx file as\n"
   + "created by a classifier DataWriter\n", programName);
 // usage message for wrong number of arguments
 if (args.length < 1) {
  System.err.format("error: wrong number of arguments\n%s", usage);
  System.exit(1);
 }
 // parse out the training directory from the arguments
 File dir = new File(args[0]);
 String[] remainingArgs = new String[args.length - 1];
 System.arraycopy(args, 1, remainingArgs, 0, remainingArgs.length);
 // train and package the classifier
 Train.main(dir, remainingArgs);
}

@Override
public void trainClassifier(File dir, String... args) throws Exception {
 String[] delegatedArgs = new String[args.length + 1];
 System.arraycopy(args, 0, delegatedArgs, 1, args.length);
 delegatedArgs[0] = this.getDelegatedModelDirectory(dir).getPath();
 Train.main(delegatedArgs);
}

public void train(File directory, Model.Params params) throws Exception {
 Train.main(this.getModelDirectory(directory, params), params.trainingArguments);
}

@Override
public void train(CollectionReader collectionReader, File outputDirectory) throws Exception {
 // assemble the training pipeline
 AggregateBuilder aggregate = new AggregateBuilder();
 // an annotator that loads the text from the training file URIs
 aggregate.add(UriToDocumentTextAnnotator.getDescription());
 // an annotator that parses and loads MASC named entity annotations (and tokens)
 aggregate.add(MascGoldAnnotator.getDescription());
 // an annotator that adds part-of-speech tags
 aggregate.add(PosTaggerAnnotator.getDescription());
 // our NamedEntityChunker annotator, configured to write Mallet CRF training data
 aggregate.add(AnalysisEngineFactory.createEngineDescription(
   NamedEntityChunker.class,
   CleartkSequenceAnnotator.PARAM_IS_TRAINING,
   true,
   DirectoryDataWriterFactory.PARAM_OUTPUT_DIRECTORY,
   outputDirectory,
   DefaultSequenceDataWriterFactory.PARAM_DATA_WRITER_CLASS_NAME,
   MalletCrfStringOutcomeDataWriter.class));
 // run the pipeline over the training corpus
 SimplePipeline.runPipeline(collectionReader, aggregate.createAggregateDescription());
 // quiet Mallet down a bit (but still leave likelihoods so you can see progress)
 Logger malletLogger = Logger.getLogger("cc.mallet");
 malletLogger.setLevel(Level.WARNING);
 Logger likelihoodLogger = Logger.getLogger("cc.mallet.fst.CRFOptimizableByLabelLikelihood");
 likelihoodLogger.setLevel(Level.INFO);
 // train a Mallet CRF model on the training data
 Train.main(outputDirectory);
}

Train.main(outputDirectory);

org.cleartk.ml.jar.Train.main(args);
logger.info("finished training.");
hider.restoreOutput();

org.cleartk.ml.jar.Train.main(args);
logger.info("finished training.");
hider.restoreOutput();

public static void main(String[] args) throws Exception {
 Options options = CliFactory.parseArguments(Options.class, args);
 // a reader that loads the URIs of the training files
 CollectionReaderDescription reader = UriCollectionReader.getDescriptionFromDirectory(
   options.getTrainDirectory(),
   MascTextFileFilter.class,
   null);
 // assemble the training pipeline
 AggregateBuilder aggregate = new AggregateBuilder();
 // an annotator that loads the text from the training file URIs
 aggregate.add(UriToDocumentTextAnnotator.getDescription());
 // an annotator that parses and loads MASC named entity annotations (and tokens)
 aggregate.add(MascGoldAnnotator.getDescription());
 // an annotator that adds part-of-speech tags (so we can use them for features)
 aggregate.add(PosTaggerAnnotator.getDescription());
 // our NamedEntityChunker annotator, configured to write Mallet CRF training data
 aggregate.add(AnalysisEngineFactory.createEngineDescription(
   NamedEntityChunker.class,
   CleartkSequenceAnnotator.PARAM_IS_TRAINING,
   true,
   DirectoryDataWriterFactory.PARAM_OUTPUT_DIRECTORY,
   options.getModelDirectory(),
   DefaultSequenceDataWriterFactory.PARAM_DATA_WRITER_CLASS_NAME,
   MalletCrfStringOutcomeDataWriter.class));
 // run the pipeline over the training corpus
 SimplePipeline.runPipeline(reader, aggregate.createAggregateDescription());
 // train a Mallet CRF model on the training data
 Train.main(options.getModelDirectory());
}