默认加载的是新浪新闻数据集，对数据集进行数据清洗、分词、gram 化处理等， 以满足 LDA 算法的输入需求，数据处理示例如下： 原始数据： 马晓旭意外受伤让国奥警惕 无奈大雨格外青睐殷家军记者傅亚雨沈阳报道 来到沈阳，国奥队依然没有摆脱雨水的困扰。

KMeans 文本聚类算法

KMeans Clusterer Clusters the points into k clusters by minimizing the total intra-cluster variance. It uses a given {@link Metric} to find the dista

Class for Arbitrary CRFs. These are CRFs with completely arbitrary graphical structure. The user passes in a list of instances of ACRF.CliqueFacto

Created: Sun Jan 25 23:28:45 2004 @author <a href="mailto:casutton@cs.umass.edu">Charles Sutton</a> @version $Id: ACRFEvaluator.java,v 1.1 2007/10/22

Created: Mar 1, 2005 @author <A HREF="mailto:casutton@cs.umass.edu>casutton@cs.umass.edu</A> @version $Id: ACRFExtractor.java,v 1.1 2007/10/22 21:38:

Created: Mar 31, 2005 @author <A HREF="mailto:casutton@cs.umass.edu>casutton@cs.umass.edu</A> @version $Id: ACRFExtractorTrainer.java,v 1.1 2007/10/2

$Id: ACRFTrainer.java,v 1.1 2007/10/22 21:37:43 mccallum Exp $

Created by IntelliJ IDEA. User: pereira Date: Jun 19, 2005 Time: 1:38:28 PM A search iterator over an underlying graph. The iterator returns search no

Created by IntelliJ IDEA. User: pereira Date: Jun 19, 2005 Time: 1:07:17 PM Search node in an A search.

Created by IntelliJ IDEA. User: pereira Date: Jun 20, 2005 Time: 5:16:05 PM Search state with heuristic cost-to-completion.

Created: Dec 15, 2005 @author <A HREF="mailto:casutton@cs.umass.edu>casutton@cs.umass.edu</A> @version $Id: AbstractAssignmentIterator.java,v 1.1 200

Abstract base class for umplementations of belief propagation for general factor graphs. This class manages arrays of messages, computing beliefs from

An Abstract class from which new Factor classes can be subclassed. Created: Sep 12, 2005 @author <A HREF="mailto:casutton@cs.umass.edu>casutton@cs.u

Abstract base class for inferencers. This simply throws an UnsupportedOperationException for all methods, which is useful for subclasses that want

Created: May 29, 2005 @author <A HREF="mailto:casutton@cs.umass.edu>casutton@cs.umass.edu</A> @version $Id: AbstractMessageStrategy.java,v 1.1 2007/1

Class for a multivariate multinomial distribution. <p/> Created: Mon Sep 15 17:19:24 2003 @author <a href="mailto:casutton@cs.umass.edu">Charles Sutt

Methods for calculating and displaying the accuracy v. coverage data for a Trial

Constructs Accuracy-coverage graph using confidence values to sort Fields. Created: Nov 8, 2005 @author <A HREF="mailto:culotta@cs.umass.edu>culotta

Accuracy of a clustering is (truePositive + trueNegative) / (numberPairwiseComparisons) @author culotta

Created: Aug 24, 2005 @author <A HREF="mailto:casutton@cs.umass.edu>casutton@cs.umass.edu</A> @version $Id: AcrfSerialEvaluator.java,v 1.1 2007/10/22

AdaBoost Robert E. Schapire. "The boosting approach to machine learning: An overview." In MSRI Workshop on Nonlinear Estimation and Classificat

AdaBoostM2 <p>Yoav Freund and Robert E. Schapire "Experiments with a New Boosting Algorithm" In Journal of Machine Learning: Proceedings of the 13th

This version of AdaBoost can handle multi-class problems. For binary classification, can also use <tt>AdaBoostTrainer</tt>. <p>Yoav Freund and Rober

This version of AdaBoost should be used only for binary classification. Use AdaBoost.M2 for multi-class problems. <p>Robert E. Schapire. "A decision-

This pipe uses a Classifier to label each token (i.e., using 0-th order Markov assumption), then adds the predictions as features to each token. This

A {@link Neighbor} created by merging two clusters of the original Clustering. @author "Aron Culotta" <culotta@degas.cs.umass.edu> @version 1.0 @sinc

Iterate over all pairs of Instances. @author "Aron Culotta" <culotta@degas.cs.umass.edu> @version 1.0 @since 1.0 @see NeighborIterator

Always adds edges that have not been touched, after that adds random edges.

A mapping between integers and objects where the mapping in each direction is efficient. Integers are assigned consecutively, starting at zero, as ob

An interface for objects that contain one or more Alphabets. <p> The primary kind of type checking among MALLET objects such as Instances, InstanceLis

Converts a Java array of numerical types to a FeatureVector, where the Alphabet is the data array index wrapped in an Integer object. @author

Static utility methods for arrays (like java.util.Arrays, but more useful). @author <a href="mailto:casutton@cs.umass.edu">Charles Sutton</a> @vers

An assignment to a bunch of variables. <p/> Note that outcomes are always integers. If you want them to be something else, then the Variables all hav

Iterates over the assignments to a set of variables. This is never instantiated by user code; instead, use one of the many assignmentIterator() meth

Add specified conjunctions to each instance. @author Andrew McCallum <a href="mailto:mccallum@cs.umass.edu">mccallum@cs.umass.edu</a>

Given an AugmentableFeatureVector, set those values greater than or equal to 1 to log(value)+1. This is useful when multiple counts should not be

Evaluate a Clustering using the B-Cubed evaluation metric. See Bagga & Baldwin, "Algorithms for scoring coreference chains." Unlike other metrics, th

Created: Nov 12, 2004 @author <A HREF="mailto:casutton@cs.umass.edu>casutton@cs.umass.edu</A> @version $Id: BIOTokenizationFilter.java,v 1.1 2007/10/

Created: May 30, 2005 @author <A HREF="mailto:casutton@cs.umass.edu>casutton@cs.umass.edu</A> @version $Id: BPRegionGenerator.java,v 1.1 2007/10/22 2

@author Andrew McCallum <a href="mailto:mccallum@cs.umass.edu">mccallum@cs.umass.edu</a>

Bagging Trainer. @author Andrew McCallum <a href="mailto:mccallum@cs.umass.edu">mccallum@cs.umass.edu</a>

Classification methods of BalancedWinnow algorithm. @see BalancedWinnowTrainer @author Gary Huang <a href="mailto:ghuang@cs.umass.edu">ghuang@cs.umas

An implementation of the training methods of a BalancedWinnow on-line classifier. Given a labeled instance (x, y) the algorithm computes dot(x, wi), f

$Id: BetaFactor.java,v 1.1 2007/10/22 21:37:44 mccallum Exp $

Bi-Normal Separation is a feature weighting algorithm introduced in: An Extensive Empirical Study of Feature Selection Metrics for Text Classificatio

A mapping between integers and objects where the mapping in each direction is efficient. Integers are assigned consecutively, starting at zero, as ob

A template that adds edges between adjacent nodes in a label sequence for one factor.

A factor over a continuous variable theta and binary variables <tt>var</tt>. such that <tt>phi(x|theta)<tt> is Potts. That is, for fixed theta, <tt>

A clique that uses very little time and memory based on the flyweight pattern. The owner supplies an Alphabet of vertices and a BitSet, and the cli

A factor over a continuous variable theta and binary variables <tt>var</tt>. such that <tt>phi(x|theta)<tt> is Potts. That is, for fixed theta, <tt>

A factor over a continuous variable theta and binary variables <tt>var</tt>. such that <tt>phi(x|theta)<tt> is Potts. That is, for fixed theta, <tt>

This interface is a tag indicating that the classifier attends to the InstanceList.getInstanceWeight() weights when training.

A Pipe that works like a rule list. Evaluate predicate() on each Pipe in the array; the first one that returns true, call that one ("the called Pipe"

Computes the joint of a GraphicalModel by brute-force calculation. This is exponentially slow, so it is mostly useful as a sanity check on more

This class reads through a single file, breaking each line into data and (optional) name and label fields.

For active learning, in which this trainer will select certain instances and request that the Labeler instance label them. @param trainingAndUnlabeled

Region computer where each top-level region consists of a single factor node. If the model is pairwise, this is equivalent to using the Bethe free en

For various kinds of online learning by batches, where training instances are presented, consumed for learning immediately. The same instances may be