/// <exception cref="System.IO.IOException"/>
/// <exception cref="System.TypeLoadException"/>
public static void Main(string[] args)
{
RedwoodConfiguration.Standard().Apply();
// Disable SLF4J crap.
ArgumentParser.FillOptions(typeof(Edu.Stanford.Nlp.IE.KBPEnsembleExtractor), args);
object @object = IOUtils.ReadObjectFromURLOrClasspathOrFileSystem(StatisticalModel);
IKBPRelationExtractor statisticalExtractor;
if (@object is LinearClassifier)
{
//noinspection unchecked
statisticalExtractor = new KBPStatisticalExtractor((IClassifier <string, string>)@object);
}
else
{
if (@object is KBPStatisticalExtractor)
{
statisticalExtractor = (KBPStatisticalExtractor)@object;
}
else
{
throw new InvalidCastException(@object.GetType() + " cannot be cast into a " + typeof(KBPStatisticalExtractor));
}
}
logger.Info("Read statistical model from " + StatisticalModel);
IKBPRelationExtractor extractor = new Edu.Stanford.Nlp.IE.KBPEnsembleExtractor(new KBPTokensregexExtractor(TokensregexDir), new KBPSemgrexExtractor(SemgrexDir), statisticalExtractor);
IList <Pair <KBPRelationExtractor.KBPInput, string> > testExamples = IKBPRelationExtractor.ReadDataset(TestFile);
extractor.ComputeAccuracy(testExamples.Stream(), Predictions.Map(null));
}
/// <exception cref="System.IO.IOException"/>
public static void Main(string[] args)
{
RedwoodConfiguration.Standard().Apply();
// Disable SLF4J crap.
ArgumentParser.FillOptions(typeof(Edu.Stanford.Nlp.IE.KBPSemgrexExtractor), args);
Edu.Stanford.Nlp.IE.KBPSemgrexExtractor extractor = new Edu.Stanford.Nlp.IE.KBPSemgrexExtractor(Dir);
IList <Pair <KBPRelationExtractor.KBPInput, string> > testExamples = IKBPRelationExtractor.ReadDataset(TestFile);
extractor.ComputeAccuracy(testExamples.Stream(), Predictions.Map(null));
}
/// <exception cref="System.IO.IOException"/>
/// <exception cref="System.TypeLoadException"/>
public static void Main(string[] args)
{
RedwoodConfiguration.Standard().Apply();
// Disable SLF4J crap.
ArgumentParser.FillOptions(typeof(Edu.Stanford.Nlp.IE.KBPStatisticalExtractor), args);
// Fill command-line options
// Load the test (or dev) data
Redwood.Util.ForceTrack("Test data");
IList <Pair <KBPRelationExtractor.KBPInput, string> > testExamples = IKBPRelationExtractor.ReadDataset(TestFile);
log.Info("Read " + testExamples.Count + " examples");
Redwood.Util.EndTrack("Test data");
// If we can't find an existing model, train one
if (!IOUtils.ExistsInClasspathOrFileSystem(ModelFile))
{
Redwood.Util.ForceTrack("Training data");
IList <Pair <KBPRelationExtractor.KBPInput, string> > trainExamples = IKBPRelationExtractor.ReadDataset(TrainFile);
log.Info("Read " + trainExamples.Count + " examples");
log.Info(string.Empty + trainExamples.Stream().Map(null).Filter(null).Count() + " are " + KBPRelationExtractorConstants.NoRelation);
Redwood.Util.EndTrack("Training data");
// Featurize + create the dataset
Redwood.Util.ForceTrack("Creating dataset");
RVFDataset <string, string> dataset = new RVFDataset <string, string>();
AtomicInteger i = new AtomicInteger(0);
long beginTime = Runtime.CurrentTimeMillis();
trainExamples.Stream().Parallel().ForEach(null);
// This takes a while per example
trainExamples.Clear();
// Free up some memory
Redwood.Util.EndTrack("Creating dataset");
// Train the classifier
log.Info("Training classifier:");
IClassifier <string, string> classifier = TrainMultinomialClassifier(dataset, FeatureThreshold, Sigma);
dataset.Clear();
// Free up some memory
// Save the classifier
IOUtils.WriteObjectToFile(new Edu.Stanford.Nlp.IE.KBPStatisticalExtractor(classifier), ModelFile);
}
// Read either a newly-trained or pre-trained model
object model = IOUtils.ReadObjectFromURLOrClasspathOrFileSystem(ModelFile);
Edu.Stanford.Nlp.IE.KBPStatisticalExtractor classifier_1;
if (model is IClassifier)
{
//noinspection unchecked
classifier_1 = new Edu.Stanford.Nlp.IE.KBPStatisticalExtractor((IClassifier <string, string>)model);
}
else
{
classifier_1 = ((Edu.Stanford.Nlp.IE.KBPStatisticalExtractor)model);
}
// Evaluate the model
classifier_1.ComputeAccuracy(testExamples.Stream(), Predictions.Map(null));
}