Package edu.stanford.nlp.ie.crf

A package for doing inference with conditional random fields.

See:
          Description

Class Summary
CRFBiasedClassifier<IN extends CoreMap> CRFBiasedClassifier is used to adjust the precision-recall tradeoff of any CRF model implemented using CRFClassifier.
CRFClassifier<IN extends CoreMap> Class for Sequence Classification using a Conditional Random Field model.
CRFClassifier.TestSequenceModel  
CRFClassifierEvaluator<IN extends CoreMap> Evaluates CRFClassifier on a set of data - called by QNMinimizer periodically - If evalCmd is set, runs command line specified by evalCmd otherwise does evaluation internally NOTE: when running conlleval with exec on Linux, linux will first fork process by duplicating memory of current process.
CRFCliqueTree Builds a CliqueTree (an array of FactorTable) and does message passing inference along it.
CRFDatum<FEAT,LAB> The representation of Datums used internally in CRFClassifier.
CRFFeatureExporter<IN extends CoreMap> Exports CRF features for use with other programs - Usage: CRFFeatureExporter -prop -trainFile -exportFeatures - Output file is automatically gzipped/b2zipped if ending in gz/bz2 - bzip2 requires that bzip2 is availaible via command line - Currently exports features in a format that can be read by a modified crfsgd (crfsgd assumes features are gzipped) TODO: Support other formats (like crfsuite)
CRFLabel  
CRFLogConditionalObjectiveFloatFunction  
CRFLogConditionalObjectiveFunction  
FactorTable Stores a factor table as a one dimensional array of doubles.
FloatFactorTable Stores a factor table as a one dimensional array of floats.
NERGUI A GUI for Named Entity sequence classifiers.
 

Package edu.stanford.nlp.ie.crf Description

A package for doing inference with conditional random fields. This implements the common, standard case of a linear chain CRF of arbitrary order (though usually first order in practice) where each position can be labeled with one of a fixed set of classes.

Through the use of different edu.stanford.nlp.ie.FeatureFactory classes and different edu.stanford.nlp.sequences.DocumentReaderAndWriter classes, it can read data in various formats, and be customized for various sequence inference tasks. Most of its use has been for Named Entity Recognition, but it has also been used for other applications such as Chinese word segmentation and OCR.

For more usage information, consult the Javadoc of the CRFClassifier class.

Author:
Jenny Finkel


Stanford NLP Group