TEES.ExampleBuilders.FeatureBuilders.MultiEdgeFeatureBuilder.MultiEdgeFeatureBuilder

Class MultiEdgeFeatureBuilder

This feature builder generates features describing a pair of word tokens connected by one or more dependencies. Most of the features it produces are built on the shortest undirected path of dependencies between the two tokens.

Instance Methods

[hide private]

__init__(self, featureSet, style=None)

source code

getEdgeType(self, edge)

source code

definePredictedValueRange(self, sentences, elementName)

source code

setFeatureVector(self, features=None, entity1=None, entity2=None, resetCache=True)
When the feature builder builds features, they are put to this feature vector.

source code

buildPredictedValueFeatures(self, element, tag)
Edge examples are usually predicted on top of predicted entities.

source code

buildEntityFeatures(self, sentenceGraph)
Build features for the two entities of the current example.

source code

buildPathLengthFeatures(self, pathTokens)
Simple numeric features about the length of the path

source code

buildSentenceFeatures(self, sentenceGraph)

source code

buildTerminusTokenFeatures(self, pathTokens, sentenceGraph)
Token features for the first and last tokens of the path

source code

buildWalkPaths(self, pathTokens, walks, sentenceGraph)

source code

buildPathGrams(self, length, pathTokens, sentenceGraph)
Goes through all the possible walks and builds features for subsections of "length" edges.

source code

addType(self, token, sentenceGraph, prefix='annType_') source code

buildPathEdgeFeatures(self, pathTokens, sentenceGraph)

source code

buildSingleElementFeatures(self, pathTokens, sentenceGraph)

source code

buildTerminusFeatures(self, token, ignoreEdges, prefix, sentenceGraph)

source code

Inherited from FeatureBuilder.FeatureBuilder: getEntityType, getPOSSuperType, getTokenAnnotatedType, getTokenFeatures, normalizeFeatureVector, setFeature, setTag

init(self, featureSet, style=None)
(Constructor)

source code

Parameters:

featureSet (IdSet) - feature ids

Overrides: FeatureBuilder.FeatureBuilder.__init__

setFeatureVector(self, features=None, entity1=None, entity2=None, resetCache=True)

source code

When the feature builder builds features, they are put to this feature vector.

Parameters:

features (dictionary) - a reference to the feature vector
entity1 (cElementTree.Element) - an entity used by trigger or edge feature builders
entity2 (cElementTree.Element) - an entity used by trigger or edge feature builders
resetCache (boolean) - Some intermediate features are cached to speed up example generation. This cache should be cleared when moving to another example.

Overrides: FeatureBuilder.FeatureBuilder.setFeatureVector

buildPredictedValueFeatures(self, element, tag)

source code

Edge examples are usually predicted on top of predicted entities. The entities' confidence scores can be used as features for edge detection. For these features to be used, the model must also have been trained on data that contains prediction confidence scores.

buildEntityFeatures(self, sentenceGraph)

source code

Build features for the two entities of the current example. These features are labeled as "e1" or "e2", so entity order is meaningful.

Class MultiEdgeFeatureBuilder

__init__(self, featureSet, style=None) (Constructor)

setFeatureVector(self, features=None, entity1=None, entity2=None, resetCache=True)

buildPredictedValueFeatures(self, element, tag)

buildEntityFeatures(self, sentenceGraph)

init(self, featureSet, style=None)
(Constructor)