DKPro TC

Discriminators (0.5.0 Release)

Mandatory Parameters

  • readers (DimensionBundle)
    • readerTrain (name of the reader for the training data)
    • readerTrainParams (further parameters for the training data reader)
    • readerTest (name of the reader for the test data) not necessary for Crossvalidation
    • readerTestParams (further parameters for the test data reader) not necessary for Crossvalidation
  • dataWriter (a writer to produce the input for the classification framework, e.g. Weka)
  • classificationArguments (the classification algorithm and a list of arguments to parametrize it)
  • featureSet (the names of the feature extractors to use)
  • pipelineParameters (further parameters necessary to configure the feature extractors)

Optional Parameters with default values

  • multiLabel: false (if true, you need to set threshold)
  • threshold: 0 needs to be set when multiLabel is true
  • featureSelection (DimensionBundle)
    • attributeEvaluator
    • featureSearcher (only for single-label)
    • labelTransformationMethod (only for multi-label)
    • numLabelsToKeep (only for multi-label)
    • applySelection
  • isRegressionExperiment: false
  • isPairClassification: false
  • isUnitClassification: false