更新2.1版本

fnlp-app/pom.xml

@@ -4,11 +4,11 @@
 	<parent>
 		<groupId>org.fnlp</groupId>
 		<artifactId>fnlp-all</artifactId>
-		<version>2.0-SNAPSHOT</version>
+		<version>2.1-SNAPSHOT</version>
 	</parent>
 	<groupId>org.fnlp</groupId>
 	<artifactId>fnlp-app</artifactId>
-	<version>2.0-SNAPSHOT</version>
+	<version>2.1-SNAPSHOT</version>
 	<name>fnlp-app</name>
 	<url>http://maven.apache.org</url>
 	<properties>
@@ -24,7 +24,7 @@
 		<dependency>
 			<groupId>org.fnlp</groupId>
 			<artifactId>fnlp-core</artifactId>
-			<version>2.0-SNAPSHOT</version>
+			<version>2.1-SNAPSHOT</version>
 		</dependency>
 		<dependency>
 			<groupId>org.apache.lucene</groupId>

fnlp-core/.classpath

@@ -20,7 +20,6 @@
 	<classpathentry kind="con" path="org.eclipse.m2e.MAVEN2_CLASSPATH_CONTAINER">
 		<attributes>
 			<attribute name="maven.pomderived" value="true"/>
-			<attribute name="org.eclipse.jst.component.nondependency" value=""/>
 		</attributes>
 	</classpathentry>
 	<classpathentry kind="output" path="target/classes"/>

fnlp-core/.project

@@ -5,11 +5,6 @@
 	<projects>
 	</projects>
 	<buildSpec>
-		<buildCommand>
-			<name>org.eclipse.wst.common.project.facet.core.builder</name>
-			<arguments>
-			</arguments>
-		</buildCommand>
 		<buildCommand>
 			<name>org.eclipse.jdt.core.javabuilder</name>
 			<arguments>
@@ -20,17 +15,9 @@
 			<arguments>
 			</arguments>
 		</buildCommand>
-		<buildCommand>
-			<name>org.eclipse.wst.validation.validationbuilder</name>
-			<arguments>
-			</arguments>
-		</buildCommand>
 	</buildSpec>
 	<natures>
-		<nature>org.eclipse.jem.workbench.JavaEMFNature</nature>
-		<nature>org.eclipse.wst.common.modulecore.ModuleCoreNature</nature>
 		<nature>org.eclipse.m2e.core.maven2Nature</nature>
 		<nature>org.eclipse.jdt.core.javanature</nature>
-		<nature>org.eclipse.wst.common.project.facet.core.nature</nature>
 	</natures>
 </projectDescription>

fnlp-core/pom.xml

@@ -4,12 +4,12 @@
 	<parent>
 		<groupId>org.fnlp</groupId>
 		<artifactId>fnlp-all</artifactId>
-		<version>2.0-SNAPSHOT</version>
+		<version>2.1-SNAPSHOT</version>
 		<relativePath>../pom.xml</relativePath>
 	</parent>
 	<groupId>org.fnlp</groupId>
 	<artifactId>fnlp-core</artifactId>
-	<version>2.0-SNAPSHOT</version>
+	<version>2.1-SNAPSHOT</version>
 	<name>fnlp-core</name>
 	<url>https://github.com/xpqiu/fnlp/</url>
 	<description>FNLP is developed for Chinese natural language processing (NLP), which also includes some machine learning algorithms and data sets to achieve the NLP tasks. FNLP is distributed under LGPL3.0.</description>

fnlp-core/src/main/java/org/fnlp/ml/classifier/AbstractClassifier.java

@@ -36,6 +36,7 @@
 public abstract class AbstractClassifier implements Serializable{
 
 	private static final long serialVersionUID = -175929257288466023L;
+
 	protected AlphabetFactory factory;
 
 

fnlp-core/src/main/java/org/fnlp/ml/classifier/linear/Linear.java

@@ -51,9 +51,7 @@ public class Linear extends AbstractClassifier implements Serializable	{
 
 	private static final long serialVersionUID = -2626247109469506636L;
 
-	protected Inferencer inferencer;
-
-	protected AlphabetFactory factory;
+	protected Inferencer inferencer;
 
 	protected Pipe pipe;
 
@@ -145,8 +143,6 @@ public Pipe getPipe() {
 		return pipe;		
 	}
 
-	public AlphabetFactory getAlphabetFactory() {
-		return factory;
-	}
+
 
 }

fnlp-core/src/main/java/org/fnlp/ml/classifier/linear/OnlineTrainer.java

@@ -40,7 +40,7 @@
 /**
  * 在线参数训练类，
  * 可能问题：收敛控制，参数c设置过小，可能会导致“假收敛”的情况 2012.8.6
- *
+ * @author xpqiu
  */
 public class OnlineTrainer extends AbstractTrainer {
 
@@ -61,7 +61,6 @@ public class OnlineTrainer extends AbstractTrainer {
 	public boolean interim = false;
 
 	public float c=0.1f;
-
 	public float threshold = 0.99f;
 
 	protected Linear classifier;
@@ -74,55 +73,57 @@ public class OnlineTrainer extends AbstractTrainer {
 	protected float[] weights;
 	AlphabetFactory af;
 
-	public OnlineTrainer(AlphabetFactory af, int iternum) {
-		//默认特征生成器
-		Generator gen = new SFGenerator();
-		//默认推理器
-		this.inferencer = new LinearMax(gen, af.getLabelSize());
-		//默认损失函数
-		this.loss =  new ZeroOneLoss();
-		//默认参数更新策略
-		this.update = new LinearMaxPAUpdate(loss);
-		this.iternum = iternum;
-		this.c = 0.1f;
-		this.af = af;
-		weights = (float[]) inferencer.getWeights();		
-		if (weights == null) {
-			weights = new float[af.getFeatureSize()];
-			inferencer.setWeights(weights);
-		}
-		random = new Random(1l);
-	}
+
 	/**
 	 * 构造函数
 	 * @param af 字典
 	 */
 	public OnlineTrainer(AlphabetFactory af) {
 		this(af,50);
+	}
+
+	public OnlineTrainer(AlphabetFactory af, int iternum) {
+		//默认特征生成器
+		Generator gen = new SFGenerator();
+		//默认推理器
+		this.inferencer = new LinearMax(gen, af.getLabelSize());
+		//默认损失函数
+		this.loss =  new ZeroOneLoss();
+		//默认参数更新策略
+		this.update = new LinearMaxPAUpdate(loss);
+		this.iternum = iternum;
+		this.af = af;
+		weights = (float[]) inferencer.getWeights();		
+		if (weights == null) {
+			weights = new float[af.getFeatureSize()];
+			inferencer.setWeights(weights);
+		}
+		random = new Random(1l);		
 	}
 
 	/**
 	 * 构造函数
 	 * @param inferencer 推理算法
 	 * @param update 参数更新方法
 	 * @param loss 损失计算方法
-	 * @param fsize 特征数量
+	 * @param af 特征标签词典
 	 * @param iternum 最大迭代次数
 	 * @param c 步长阈值
 	 */
 	public OnlineTrainer(Inferencer inferencer, Update update,
-			Loss loss, int fsize, int iternum, float c) {
+			Loss loss, AlphabetFactory af, int iternum, float c) {
 		this.inferencer = inferencer;
 		this.update = update;
 		this.loss = loss;
 		this.iternum = iternum;
-		this.c = c;
+		this.c = c;
+		this.af =af;
 		weights = (float[]) inferencer.getWeights();
 		if (weights == null) {
-			weights = new float[fsize];
+			weights = new float[af.getFeatureSize()];
 			inferencer.setWeights(weights);
-		}else if(weights.length<fsize){
-			weights = Arrays.copyOf(weights, fsize);
+		}else if(weights.length<af.getFeatureSize()){
+			weights = Arrays.copyOf(weights, af.getFeatureSize());
 			inferencer.setWeights(weights);
 		}
 		random = new Random(1l);
@@ -133,12 +134,12 @@ public OnlineTrainer(Inferencer inferencer, Update update,
 	 * @param classifier 分类器
 	 * @param update 参数更新方法
 	 * @param loss 损失计算方法
-	 * @param fsize 特征数量
+	 * @param af  特征标签词典
 	 * @param iternum 最大迭代次数
 	 * @param c 步长阈值
 	 */
-	public OnlineTrainer(Linear classifier, Update update, Loss loss, int fsize, int iternum, float c) {
-		this(classifier.getInferencer(), update, loss, fsize, iternum, c);
+	public OnlineTrainer(Linear classifier, Update update, Loss loss, AlphabetFactory af, int iternum, float c) {
+		this(classifier.getInferencer(), update, loss, af, iternum, c);
 	}
 
 	/**
@@ -252,7 +253,7 @@ public Linear train(InstanceSet trainset, InstanceSet devset) {
 			System.out.println();			
 
 			if (interim) {
-				Linear p = new Linear(inferencer, trainset.getAlphabetFactory());
+				Linear p = new Linear(inferencer, af);
 				try {
 					p.saveTo("tmp.model");
 				} catch (IOException e) {

fnlp-core/src/main/java/org/fnlp/nlp/cn/PartOfSpeech.java

@@ -28,7 +28,8 @@ public enum PartOfSpeech {
 	专有名,
 	人名,
 	地名,
-	机构名,
+	机构名,
+	实体名,
 	型号名,
     事件名,
     网址,
@@ -41,7 +42,7 @@ public enum PartOfSpeech {
 
 	代词,
 	人称代词, 
-	指示代词, 
+	指示词, 
 	疑问代词,
 
 
@@ -58,8 +59,8 @@ public enum PartOfSpeech {
 
 
     动词,
-    能愿动词,
-    趋向动词,
+    情态词,
+    趋向词,
     被动词,
     把动词,
 
@@ -104,7 +105,7 @@ public static PartOfSpeech[] valueOf(String[] pos) {
 	static{
 		Pronoun.add(代词);
 		Pronoun.add(人称代词);
-		Pronoun.add(指示代词);
+		Pronoun.add(指示词);
 		Pronoun.add(疑问代词);
 	}
 	/**
@@ -129,7 +130,8 @@ public boolean isMark() {
 		entities.add(人名);
 		entities.add(地名);
 		entities.add(机构名);
-		entities.add(专有名);
+		entities.add(专有名);
+		entities.add(实体名);
 	}
 
 	/**
@@ -141,13 +143,18 @@ public boolean isEntiry() {
 		return entities.contains(this);
 	}
 	/**
-	 * 判断词性是否为一个实体，包括：人名|地名|机构名|专有名。
+	 * 判断词性是否为一个实体，包括：人名|地名|机构名|专有名|实体名。
 	 * @param pos
 	 * @return
 	 */
 	public static boolean isEntiry(String pos) {
-		PartOfSpeech p = valueOf(pos);
-
+		PartOfSpeech p;
+		try {
+			p = valueOf(pos);
+		} catch (Exception e) {
+			System.err.println(pos+"不存在");
+			return false;
+		}		
 		return p.isEntiry();
 	}
 

fnlp-core/src/main/java/org/fnlp/nlp/cn/anaphora/EntitiesGetter.java

@@ -49,7 +49,7 @@ public class EntitiesGetter {
 		NP.add(PartOfSpeech.限定词);
 		NP.add(PartOfSpeech.名词);
 		NP.add(PartOfSpeech.代词);
-		NP.add(PartOfSpeech.指示代词);
+		NP.add(PartOfSpeech.指示词);
 		NP.add(PartOfSpeech.人称代词);
 		NP.add(PartOfSpeech.疑问代词);
 
@@ -218,7 +218,7 @@ else if(this.isMale(ey.getData())){
 		NN.add(PartOfSpeech.机构名);
 		NN.add(PartOfSpeech.代词);
 		NN.add(PartOfSpeech.人称代词);
-		NN.add(PartOfSpeech.指示代词);
+		NN.add(PartOfSpeech.指示词);
 		NN.add(PartOfSpeech.疑问代词);
 	}
 

fnlp-core/src/main/java/org/fnlp/nlp/cn/anaphora/train/ARClassifier.java

@@ -146,7 +146,7 @@ public void train() throws Exception {
 			ZeroOneLoss l = new ZeroOneLoss();
 			Inferencer ms = new LinearMax(gen, factory.getLabelSize());
 			Update update = new LinearMaxPAUpdate(l);
-			OnlineTrainer trainer = new OnlineTrainer(ms, update,l, factory.getFeatureSize(), 50,0.005f);
+			OnlineTrainer trainer = new OnlineTrainer(ms, update,l, factory, 50,0.005f);
 			Linear pclassifier = trainer.train(instset,instset);
 			pipe.removeTargetPipe();
 			pclassifier.setPipe(pipe);

fnlp-core/src/main/java/org/fnlp/nlp/parser/dep/train/JointParerTrainer.java

@@ -195,10 +195,8 @@ private InstanceSet buildInstanceList(String file) throws IOException {
 	public void train(String dataFile, int maxite, float c) throws IOException {
 
 		InstanceSet instset =  buildInstanceList(dataFile);
-		IFeatureAlphabet features = factory.DefaultFeatureAlphabet();
 
 		SFGenerator generator = new SFGenerator();
-		int fsize = features.size();
 
 		LabelAlphabet la = factory.DefaultLabelAlphabet();
 		int ysize = la.size();
@@ -207,7 +205,7 @@ public void train(String dataFile, int maxite, float c) throws IOException {
 		ZeroOneLoss loss = new ZeroOneLoss();
 		Update update = new LinearMaxPAUpdate(loss);
 		OnlineTrainer trainer = new OnlineTrainer(solver, update, loss,
-				fsize, maxite, c);
+				factory, maxite, c);
 		Linear models = trainer.train(instset, null);
 		instset = null;
 		solver = null;

fnlp-core/src/main/java/org/fnlp/nlp/parser/dep/train/ParserTrainer.java

@@ -191,11 +191,9 @@ public void train(String dataFile, int maxite, float c) throws IOException {
 
 		LabelAlphabet postagAlphabet = factory.buildLabelAlphabet("postag");
 
-		IFeatureAlphabet features = factory.DefaultFeatureAlphabet();
 
 		SFGenerator generator = new SFGenerator();
 		Linear[] models = new Linear[postagAlphabet.size()];
-		int fsize = features.size();
 
 		for (int i = 0; i < postagAlphabet.size(); i++) {
 			String pos = postagAlphabet.lookupString(i);
@@ -208,7 +206,7 @@ public void train(String dataFile, int maxite, float c) throws IOException {
 			ZeroOneLoss loss = new ZeroOneLoss();
 			Update update = new LinearMaxPAUpdate(loss);
 			OnlineTrainer trainer = new OnlineTrainer(solver, update, loss,
-					fsize, maxite, c);
+					factory, maxite, c);
 			models[i] = trainer.train(instset, null);
 			instset = null;
 			solver = null;

fnlp-core/src/main/java/org/fnlp/nlp/tag/AbstractTagger.java

@@ -92,10 +92,10 @@ public void train() throws Exception {
 		OnlineTrainer trainer;
 
 		if(cl!=null){
-			trainer = new OnlineTrainer(cl, update, loss, features.size(),iterNum, c);
+			trainer = new OnlineTrainer(cl, update, loss, factory,iterNum, c);
 		}else{
 			trainer = new OnlineTrainer(inference, update, loss,
-					features.size(), iterNum, c);
+					factory, iterNum, c);
 		}
 
 		cl = trainer.train(trainSet, testSet);