Package cc.mallet.fst

Transducers, including Conditional Random Fields (CRFs).


Interface Summary
CacheStaleIndicator Indicates when the value/gradient during training becomes stale.
MaxLattice The interface to classes implementing the Viterbi algorithm, finding the best sequence of states for a given input sequence.
SumLattice Interface to perform forward-backward during training of a transducer.
Transducer.Incrementor Methods to be called by inference methods to indicate partial counts of sufficient statistics.

Class Summary
CRF Represents a CRF model.
CRF.Factors A simple, transparent container to hold the parameters or sufficient statistics for the CRF.
CRFCacheStaleIndicator Indicates when the value/gradient becomes stale based on updates to CRF's parameters.
CRFOptimizableByBatchLabelLikelihood Implements label likelihood gradient computations for batches of data, can be easily parallelized.
CRFOptimizableByGradientValues A CRF objective function that is the sum of multiple objective functions that implement Optimizable.ByGradientValue.
CRFOptimizableByLabelLikelihood An objective function for CRFs that is the label likelihood plus a Gaussian or hyperbolic prior on parameters.
CRFTrainerByL1LabelLikelihood CRF trainer that implements L1-regularization.
CRFTrainerByLabelLikelihood Unlike ClassifierTrainer, TransducerTrainer is not "stateless" between calls to train.
CRFTrainerByStochasticGradient Trains CRF by stochastic gradient.
CRFTrainerByValueGradients A CRF trainer that can combine multiple objective functions, each represented by a Optmizable.ByValueGradient.
CRFWriter Saves a trained model to specified filename.
HMM A Hidden Markov Model.
InstanceAccuracyEvaluator Reports the percentage of instances for which the entire predicted sequence was correct.
LabelDistributionEvaluator Prints predicted and true label distribution.
MaxLatticeDefault Default, full dynamic programming version of the Viterbi "Max-(Product)-Lattice" algorithm.
MEMM A Maximum Entropy Markov Model.
MEMMTrainer Trains and evaluates a MEMM.
MultiSegmentationEvaluator Evaluates a transducer model, computes the precision, recall and F1 scores; considers segments that span across multiple tokens.
NoopTransducerTrainer A TransducerTrainer that does no training, but simply acts as a container for a Transducer; for use in situations that require a TransducerTrainer, such as the TransducerEvaluator methods.
PerClassAccuracyEvaluator Determines the precision, recall and F1 on a per-class basis.
Segment Represents a labelled chunk of a Sequence segmented by a Transducer, usually corresponding to some object extracted from an input Sequence.
ShallowTransducerTrainer Deprecated. Use NoopTransducerTrainer instead
SimpleTagger This class's main method trains, tests, or runs a generic CRF-based sequence tagger.
SimpleTagger.SimpleTaggerSentence2FeatureVectorSequence Converts an external encoding of a sequence of elements with binary features to a FeatureVectorSequence.
SumLatticeDefault Default, full dynamic programming implementation of the Forward-Backward "Sum-(Product)-Lattice" algorithm
SumLatticeFactory Provides factory methods to create inference engine for training a transducer.
ThreadedOptimizable An adaptor for optimizables based on batch values/gradients.
TokenAccuracyEvaluator Evaluates a transducer model based on predictions of individual tokens.
Transducer A base class for all sequence models, analogous to classify.Classifier.
Transducer.State An abstract class used to represent the states of the transducer.
Transducer.TransitionIterator An abstract class to iterate over the states of the transducer.
TransducerEvaluator An abstract class to evaluate a transducer model.
TransducerTrainer An abstract class to train and evaluate a transducer model.
ViterbiWriter Prints the input instances along with the features and the true and predicted labels to a file.

Package cc.mallet.fst Description

Transducers, including Conditional Random Fields (CRFs).