CollectionReaderDescription/AnalysisEngineDescription as Discriminator

[Torsten Zesch]

When using CollectionReaderDescription/AnalysisEngineDescription as Discriminators the values in the discriminators.txt are somewhat bloated as the Descriptors seem to contain whole type systems.

Is there a way around that, e.g. by "naming" discriminator values?

thanks,

Torsten

[Richard Eckart de Castilho]

Hi,

if you use Groovy, yes ;) But the support is in general also available for Java.

So first Groovy ;)

def dimDataset = new ClosureDimension("dataset", [
datasetX: { createReaderDescription(...) },
datasetY: { createReaderDescription(...) }])

The discriminators.txt file will then have

dataset = datasetX

The underlying support to enabled this is the "Discriminable" and has
e.g. also been used in DKPro TC to name feature extractor configurations:

de.tudarmstadt.ukp.dkpro.tc.core.util.ExperimentUtil.NamedArrayList<T>

Cheers,

-- Richard