bug fix

Author: xpqiu

fnlp-core/src/main/java/org/fnlp/data/reader/Reader.java

@@ -22,6 +22,7 @@
 import java.util.Iterator;
 
 import org.fnlp.ml.types.Instance;
+import org.fnlp.ml.types.InstanceSet;
 
 /**
  * @author xpqiu
@@ -35,4 +36,16 @@ public abstract class Reader implements Iterator<Instance> {
 	public void remove () {
 		throw new IllegalStateException ("This Iterator<Instance> does not support remove().");
 	}
+	
+	
+	public InstanceSet read(){
+		InstanceSet instSet = new InstanceSet();
+		while (hasNext()) {
+            Instance inst = next();
+            if(inst!=null){
+            	instSet.add(inst);
+            }
+		}
+		return instSet;
+	}
 }

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

@@ -24,6 +24,7 @@
 import org.fnlp.ml.classifier.LabelParser.Type;
 import org.fnlp.ml.types.Instance;
 import org.fnlp.ml.types.InstanceSet;
+import org.fnlp.ml.types.alphabet.AlphabetFactory;
 
 /**
  * 分类器抽象类
@@ -35,7 +36,12 @@
 public abstract class AbstractClassifier implements Serializable{
 
 	private static final long serialVersionUID = -175929257288466023L;
-
+	protected AlphabetFactory factory;
+	
+	
+	public AlphabetFactory getAlphabetFactory() {
+		return factory;
+	}
 
 	/**
 	 * 返回分类内部结果，标签为内部表示索引，需要还原处理

fnlp-core/src/main/java/org/fnlp/ml/classifier/bayes/BayesClassifier.java

@@ -31,7 +31,7 @@
  *
  */
 public class BayesClassifier extends AbstractClassifier implements Serializable{
-	protected AlphabetFactory factory;
+	
 	protected ItemFrequency tf;
 	protected Pipe pipe;
 	protected FeatureSelect fs;

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

@@ -52,7 +52,7 @@ public class Linear extends AbstractClassifier implements Serializable	{
 	private static final long serialVersionUID = -2626247109469506636L;
 
 	protected Inferencer inferencer;
-	protected AlphabetFactory factory;
+	
 	protected Pipe pipe;
 
 	public Linear(Inferencer inferencer, AlphabetFactory factory) {
@@ -126,9 +126,7 @@ public void setInferencer(Inferencer inferencer)	{
 		this.inferencer = inferencer;
 	}
 
-	public AlphabetFactory getAlphabetFactory() {
-		return factory;
-	}
+	
 
 	public void setWeights(float[] weights) {
 		inferencer.setWeights(weights);

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

@@ -71,7 +71,8 @@ public class OnlineTrainer extends AbstractTrainer {
 	protected Random random;
 
 	public int iternum;
-	protected float[] weights;
+	protected float[] weights;
+	AlphabetFactory af;
 
 	public OnlineTrainer(AlphabetFactory af, int iternum) {
 		//默认特征生成器
@@ -84,7 +85,8 @@ public OnlineTrainer(AlphabetFactory af, int iternum) {
 		this.update = new LinearMaxPAUpdate(loss);
 		this.iternum = iternum;
 		this.c = 0.1f;
-		weights = (float[]) inferencer.getWeights();
+		this.af = af;
+		weights = (float[]) inferencer.getWeights();		
 		if (weights == null) {
 			weights = new float[af.getFeatureSize()];
 			inferencer.setWeights(weights);
@@ -283,7 +285,7 @@ public Linear train(InstanceSet trainset, InstanceSet devset) {
 		System.out.println("time escape:" + (endTime - beginTime) / 1000.0
 				+ "s");
 		System.out.println();
-		Linear p = new Linear(inferencer, trainset.getAlphabetFactory());
+		Linear p = new Linear(inferencer,af);
 		return p;
 	}
 

fnlp-core/src/main/java/org/fnlp/ml/eval/Evaluation.java

@@ -50,8 +50,7 @@ public Evaluation(InstanceSet test) {
 		for(int i=0;i<totnum;i++){
 			golden[i] = (Integer) test.getInstance(i).getTarget();
 		}
-		labels = test.getAlphabetFactory().DefaultLabelAlphabet();
-		numofclass = labels.size();
+		
 
 	}
 
@@ -82,6 +81,8 @@ public void eval2File(AbstractClassifier cl,String path){
 	 * @return
 	 */
 	public void eval(AbstractClassifier cl,int nbest){
+		labels = cl.getAlphabetFactory().DefaultLabelAlphabet();
+		numofclass = labels.size();
 		TPredict[] pred = cl.classify(test,nbest);
 		int[] acc = new int[nbest];
 		for(int i=0;i<totnum;i++){

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

@@ -246,13 +246,13 @@ private void toList(ArrayList<List<String>> lists) {
 		}
 
 	}
-	
+	//错误
 	public String[] getWords() {
 		String[] words = new String[size];
 		getWords(words);
 		return words;
 	}
-
+	//错误
 	private void getWords(String[] words) {
 		words[id] = word;
 		for (int i = 0; i < leftChilds.size(); i++) {
@@ -262,6 +262,29 @@ private void getWords(String[] words) {
 			rightChilds.get(i).getWords(words);
 		}
 
+	}
+	/**
+	 * 得到依赖类型字符串
+	 * @return
+	 */
+		
+	public String getTypes()    {
+		StringBuffer sb = new StringBuffer();
+		String ste;
+		String[] str;
+		for (int i = 0; i < leftChilds.size(); i++) {
+			sb.append(leftChilds.get(i).getTypes());
+		}
+		if(relation!=null)
+			sb.append(relation);	
+		else
+			sb.append("核心词");
+		sb.append(" ");
+		for (int i = 0; i < rightChilds.size(); i++) {
+			sb.append(rightChilds.get(i).getTypes());
+		}
+				
+			return sb.toString();	
 	}
 
 }

fnlp-core/src/main/java/org/fnlp/nlp/parser/dep/analysis/AnalysisTest.java

@@ -92,8 +92,5 @@ private void print(){
 		System.out.printf("rate(UEM):\t%.8f\ttotal(sent):\t%d\n", 1 - 1.0
 				* errsent / sent_sum, sent_sum);
 	}
-	
-	
-	
 
 }

fnlp-demo/src/main/java/org/fnlp/demo/nlp/tc/TextClassificationCustom1.java

@@ -59,35 +59,36 @@ public static void main(String[] args) throws Exception {
 		AlphabetFactory af = AlphabetFactory.buildFactory();
 
 		//使用n元特征
-		Pipe ngrampp = new NGram(new int[] {2,3 });
+		Pipe ngrampp = new NGram(new int[] {1});
 		//将字符特征转换成字典索引
 		Pipe indexpp = new StringArray2IndexArray(af);
 		//将目标值对应的索引号作为类别
 		Pipe targetpp = new Target2Label(af.DefaultLabelAlphabet());		
 
 		//建立pipe组合
-		SeriesPipes pp = new SeriesPipes(new Pipe[]{targetpp,ngrampp,indexpp});
-		
-		InstanceSet trainset = new InstanceSet(pp,af);
-		
-		InstanceSet testset = new InstanceSet(pp,af);
+		SeriesPipes pp = new SeriesPipes(new Pipe[]{ngrampp,indexpp});
 
 		//用不同的Reader读取相应格式的文件
-		Reader reader = new DocumentReader(trainDataPath);
+		Reader reader = new DocumentReader(trainDataPath);		
+		InstanceSet trainset = reader.read();
 
-		//读入数据，并进行数据处理
-		trainset.loadThruStagePipes(reader);
-//		af.setStopIncrement(true);
 		reader = new DocumentReader(testDataPath);
-			
-		testset.loadThruStagePipes(reader);
+		InstanceSet testset = reader.read();
+		
+		targetpp.process(trainset);
+		targetpp.process(testset);
+		
+		pp.process(trainset);
+		af.setStopIncrement(true);
+		pp.process(testset);		
+		
 
 
 		/**
 		 * 建立分类器
 		 */		
-		OnlineTrainer trainer = new OnlineTrainer(af);
-		Linear pclassifier = trainer.train(trainset);
+		OnlineTrainer trainer = new OnlineTrainer(af,50);
+		Linear pclassifier = trainer.train(trainset,testset);
 		pp.removeTargetPipe();
 		pclassifier.setPipe(pp);
 		af.setStopIncrement(true);
@@ -101,7 +102,7 @@ public static void main(String[] args) throws Exception {
 
 		//性能评测
 		Evaluation eval = new Evaluation(testset);
-		eval.eval(cl,1);
+		eval.eval(cl,2);