module__TEESTrain

#org.bibliome.alvisnlp.modules.tees.TEESTrain

Synopsis

Train a model that predict an Alvis relation with TEES Trainer

Description

org.bibliome.alvisnlp.modules.tees.TEESTrain executes the TEES training on Corpus and record the results in Relation. Param relationName sets the name of the binary relation to predict. relationRole1 and relationRole set the two roles of the relation. Params trainSetFeature, devSetFeature and testSetFeature give respectively the features key of the train, dev and test corpus. org.bibliome.alvisnlp.modules.tees.TEESTrain

Parameters

modelTargetDir

Optional

Type: OutputFile

Path to the directory where put the trained model

namedEntityLayerName

Optional

Type: String

Name of the layer containing the named entities

schema

Optional

Type: MultiMapping

  	Give the schema of the relations to train i.e.
```xml

  	  <schema>
    <Lives_In>Bacteria,Location</Lives_In>
  	  </schema>
```

teesHome

Optional

Type: InputDirectory

Path to tees home directory.

constantRelationFeatures

Optional

Type: Mapping

UNDOCUMENTED

constantTupleFeatures

Optional

Type: Mapping

UNDOCUMENTED

corpusSetFeature

Default value: set

Type: String

UNDOCUMENTED

devSetValue

Default value: dev

Type: String

Feature key of the dev set corpus.

documentFilter

Default value: true

Type: Expression

UNDOCUMENTED

modelName

Default value: test-model

Type: String

give a name to the trained model

namedEntityTypeFeature

Default value: neType

Type: String

Name of the feature to access the type of the named entities

omitSteps

Default value: SPLIT-SENTENCES,NER

Type: String

Set the preprocessing steps to omit in the form of [SPLIT-SENTENCES][,NER][,PARSE][,FIND-HEADS]

posFeature

Default value: pos

Type: String

UNDOCUMENTED

sectionFilter

Default value: boolean:and(true, boolean:and(nav:layer:words(), nav:layer:sentences()))

Type: Expression

UNDOCUMENTED

sentenceLayerName

Default value: sentences

Type: String

UNDOCUMENTED

testSetValue

Default value: test

Type: String

Feature key of the test set corpus.

tokenLayerName

Default value: words

Type: String

UNDOCUMENTED

trainSetValue

Default value: train

Type: String

Feature key of the train set corpus.