Package edu.stanford.nlp.classify

Data Structures and Algorithms for Classification.

See:
Description

Interface Summary

Classifier
ClassifierB	An interface for a classifier.
Filter	A Filter takes a String object and returns the String after having applied the proper filter to it.
Smoother	Interface for classes which smooth a matrix
SmootherB	The Smoother should return a smoothed ProbabilitySet.
TokenReader	Title: Naive Bayes Classifier Description: An interface for converting something into a continuous string Copyright: Copyright (c) 2001 Company:

Class Summary

AbstractClassifier
AddEpsilonSmoother	Smooths a ProbabilitySet using a HashMap vocabulary.
AddOneSmoother
ButFilter	This (very primitive) filter removes everything from the beginning of a sentence until the word "but".
CategoryMatrix	Wrapper for a Matrix whose columns are P(w|c).
Classification	This class holds the classification assigned to a FeatureSet
Classify	Takes two arguments: the first is a directory where the training sets are held.
ClassProbability	This class holds the counts and probabilities for the features of a single class (classification category).
ClassProbBuilder	Accepts a feature vector and a class id, and returns a ClassProbability
CrossValidator	This class takes a probability set and does cross-validation for as many folds as desired.
Feature	A wrapper for a String which allows you to add other attributes, such as weights, to a Feature.
FeatureAdder	This class takes a ProbabilitySet and the number of classes, and returns a new ProbabilitySet where each ClassProbability represents the probabilities for the entire class.
FeatureMaker	This class accepts a string and returns a HashMap that holds the counts for each token in the string.
FeatureSet	This class stores pairs of features and counts.
FeatureSetBuilder	Given a HashMap of Features, and an ID (true classification), this class can create a FeatureSet.
FileSorter	Takes a file that contains a set of lists of double/key pairs, one to a line, and sorts them.
FirstNFilter	This filter is used if you only want to use the first n sentences in a document.
GTPreparer	This class prepares the ProbabilitySet to_be_smooted for smoothing, and calls the Good Turing smoother
ID
InfomapSmoother
KNN	K nearest-neighbors classifier.
LinkedList	A linked list class.
MassVerifier	Tracks the statistics for a set of test documents.
MedlineHandler
NaiveBayesClassifier	Naive Bayes Classifier.
NBClassifier	The classifier takes a set of (smoothed) probability vectors and a test vector, and classifies the test vector according to the probability vectors
NegativeFilter	This filter does some very primitive negative scoping, using the same method as the Almaden code.
Node	For the LinkedList?
POSFilter	A filter that takes a of tagged text string and a list of parts of speech to be kept, and returns the string that includes only those part of speech.
POSPrepFilter	Title: Description: Copyright: Copyright (c) 2001 Company:
POSTokenReader	This class reads in Epinions data that has been POS tagged using the Qtag system.
ProbabilitySet	This class manipulates an array of ClassProbabilities.
RemoveSpacesFilter	This filter removes the spaces in the middle of the POS tags, so that tokenization will happen properly.
ScoreHandler
SimpleGoodTuring
SortedArray	Maintains a sorted array of ints.
Tester	Builds a FeatureSet for the files in the directory "parent_dir" and attempts to classify them using the ProbabilitySet my_ps that was created during training.
TokenBreaker	Title: Naive Bayes Classifier Description: Puts a space between all tokens Copyright: Copyright (c) 2001 Company:
Trainer	Builds a ProbabilitySet for the classes in String[] my_dirnames and smooths that ProbabilitySet.
Verifier	Compares the classification of a FeatureSet (true value) to the classification of a Classification (predicted value)
VocabBuilder	Takes a ProbabilitySet and builds a hash table that contains the entire vocabulary.

Package edu.stanford.nlp.classify Description

Data Structures and Algorithms for Classification. This package is currently being modified. Currently implements KNN and a passable but not good version of Naive Bayes. Right now it also has a lot of stuff in there that doesn't belong (POSFilter, Smoothers, LinkedList, NegativeFilter, etc.).

Sepandar David Kamvar

Last modified: Thu Oct 31 10:56:14 PST 2002