tarexp.component.ranker module

TARexp supports classification models implemented in Scikit-learn through tarexp.component.ranker.SklearnRanker wrapper. However, any supervised learning model that can produce a score for each document in the collection can be integrated into TARexp.

Caution

For rankers that require allocating a lot of memory or carries states (e.g. neural models and SGDClassifier in sklearn that supports partial_fit), it would be ideal to put the actual model initialization into .begin method and properly dispose the model instance in .reset method.

class tarexp.component.ranker.Ranker(**kwargs)

Bases: Component

trainRanker(X, y, *args, **kwargs)

scoreDocuments(X, *args, **kwargs)

class tarexp.component.ranker.SklearnRanker(module, **kwargs)

Bases: Ranker

reset()

trainRanker(X, y, **kwargs)

scoreDocuments(X, **kwargs)

tarexp.component.ranker.addDummyNeg(X, y)