dependencypipe.java

MSTParser是以最大生成树理论为基础的判别式依存句法分析器。它将一科依存树的得分看作是 所有依存关系的得分的总和

	fv = add("JJ="+head,1.0,fv); //this
	fv = add("KK="+child+" "+childP,1.0,fv);
	fv = add("LL="+childP,1.0,fv);
	fv = add("MM="+child,1.0,fv); //this

	if(head.length() > 5 || child.length() > 5) {
	    int hL = head.length();
	    int cL = child.length();
		    
	    head = hL > 5 ? head.substring(0,5) : head;
	    child = cL > 5 ? child.substring(0,5) : child;
		    
	    all = head + " " + headP + " " + child + " " + childP;
	    hPos = headP + " " + child + " " + childP;
	    cPos = head + " " + headP + " " + childP;
	    hP = headP + " " + child;
	    cP = head + " " + childP;
	    oPos = headP + " " + childP;
	    oLex = head + " " + child;
	
	    fv = add("SA="+all+attDist,1.0,fv); //this
	    fv = add("SF="+oLex+attDist,1.0,fv); //this
	    fv = add("SAA="+all,1.0,fv); //this
	    fv = add("SFF="+oLex,1.0,fv); //this

	    if(cL > 5) {
		fv = add("SB="+hPos+attDist,1.0,fv);
		fv = add("SD="+hP+attDist,1.0,fv);
		fv = add("SK="+child+" "+childP+attDist,1.0,fv);
		fv = add("SM="+child+attDist,1.0,fv); //this
		fv = add("SBB="+hPos,1.0,fv);
		fv = add("SDD="+hP,1.0,fv);
		fv = add("SKK="+child+" "+childP,1.0,fv);
		fv = add("SMM="+child,1.0,fv); //this
	    }
	    if(hL > 5) {
		fv = add("SC="+cPos+attDist,1.0,fv);
		fv = add("SE="+cP+attDist,1.0,fv);
		fv = add("SH="+head+" "+headP+attDist,1.0,fv);
		fv = add("SJ="+head+attDist,1.0,fv); //this
			
		fv = add("SCC="+cPos,1.0,fv);
		fv = add("SEE="+cP,1.0,fv);
		fv = add("SHH="+head+" "+headP,1.0,fv);
		fv = add("SJJ="+head,1.0,fv); //this
	    }
	}
		
	return fv;
		
    }
	
    public FeatureVector createFeatureVector(String[] toks,
					     String[] pos,
					     String[] posA,
					     int word,
					     String type,
					     boolean attR,
					     boolean childFeatures,
					     FeatureVector fv) {
		
	if(!labeled) return fv;

	String att = "";
	if(attR)
	    att = "RA";
	else
	    att = "LA";

	att+="&"+childFeatures;
		
	String w = toks[word];
	String wP = pos[word];

	String wPm1 = word > 0 ? pos[word-1] : "STR";
	String wPp1 = word < pos.length-1 ? pos[word+1] : "END";

	fv = add("NTS1="+type+"&"+att,1.0,fv);
	fv = add("ANTS1="+type,1.0,fv);
	for(int i = 0; i < 2; i++) {
	    String suff = i < 1 ? "&"+att : "";
	    suff = "&"+type+suff;

	    fv = add("NTH="+w+" "+wP+suff,1.0,fv);
	    fv = add("NTI="+wP+suff,1.0,fv);
	    fv = add("NTIA="+wPm1+" "+wP+suff,1.0,fv);
	    fv = add("NTIB="+wP+" "+wPp1+suff,1.0,fv);
	    fv = add("NTIC="+wPm1+" "+wP+" "+wPp1+suff,1.0,fv);
	    fv = add("NTJ="+w+suff,1.0,fv); //this
			
	}
		
	return fv;
    }
	
    public FeatureVector createFeatureVector(String[] toks,
					     String[] pos,
					     String[] labs,
					     int[] deps) {

	String[] posA = new String[pos.length];
	for(int i = 0; i < pos.length; i++) {
	    posA[i] = pos[i].substring(0,1);
	}
		
	FeatureVector fv = new FeatureVector(-1,-1.0,null);
	for(int i = 0; i < toks.length; i++) {
	    if(deps[i] == -1)
		continue;
	    int small = i < deps[i] ? i : deps[i];
	    int large = i > deps[i] ? i : deps[i];
	    boolean attR = i < deps[i] ? false : true;
	    fv = createFeatureVector(toks,pos,posA,small,large,attR,fv);
	    if(labeled) {
		fv = createFeatureVector(toks,pos,posA,i,labs[i],attR,true,fv);
		fv = createFeatureVector(toks,pos,posA,deps[i],labs[i],attR,false,fv);
	    }
	}
	return fv;
    }

    public FeatureVector add(String feat, double val, FeatureVector fv) {
	int num = dataAlphabet.lookupIndex(feat);
	if(num >= 0)
	    return new FeatureVector(num,val,fv);
	return fv;
    }

    public void possibleFeatures(DependencyInstance inst, ObjectOutputStream out) {
	String[] toks = inst.sentence;
	String[] pos = inst.pos;
	String[] labs = inst.labs;
		
	String[] posA = new String[pos.length];
	for(int i = 0; i < pos.length; i++) {
	    posA[i] = pos[i].substring(0,1);
	}

		
	try {

	    for(int w1 = 0; w1 < toks.length; w1++) {
		for(int w2 = w1+1; w2 < toks.length; w2++) {
					
		    for(int ph = 0; ph < 2; ph++) {						
			boolean attR = ph == 0 ? true : false;

			int childInt = attR ? w2 : w1;
			int parInt = attR ? w1 : w2;
						
			FeatureVector prodFV = createFeatureVector(toks,pos,posA,w1,w2,attR,
								   new FeatureVector(-1,-1.0,null));
								
			for(FeatureVector curr = prodFV; curr != null; curr = curr.next) {
			    if(curr.index >= 0)
				out.writeInt(curr.index);
			}
			out.writeInt(-2);
								
		    }
		}
			
	    }

	    out.writeInt(-3);

	    if(labeled) {
		for(int w1 = 0; w1 < toks.length; w1++) {
		    
		    for(int t = 0; t < types.length; t++) {
			String type = types[t];
			
			for(int ph = 0; ph < 2; ph++) {						
			    boolean attR = ph == 0 ? true : false;
			    
			    for(int ch = 0; ch < 2; ch++) {						
				boolean child = ch == 0 ? true : false;						
				
				FeatureVector prodFV = createFeatureVector(toks,pos,posA,w1,
									   type,
									   attR,child,
									   new FeatureVector(-1,-1.0,null));
				
				for(FeatureVector curr = prodFV; curr != null; curr = curr.next) {
				    if(curr.index >= 0)
					out.writeInt(curr.index);
				}
				out.writeInt(-2);
				
			    }
			}
		    }
		    
		}
		
		out.writeInt(-3);
	    }

	    for(FeatureVector curr = inst.fv; curr.next != null; curr = curr.next)
		out.writeInt(curr.index);

	    out.writeInt(-4);
	    out.writeObject(inst.sentence);
	    out.writeInt(-5);
	    out.writeObject(inst.pos);
	    out.writeInt(-6);
	    out.writeObject(inst.labs);
	    out.writeInt(-7);
	    out.writeObject(inst.actParseTree);
			
	    out.writeInt(-1);
	    out.reset();

	} catch (IOException e) {}
		
    }
	
    public DependencyInstance getFeatureVector(ObjectInputStream in,
					       DependencyInstance inst,
					       FeatureVector[][][] fvs,
					       double[][][] probs,
					       FeatureVector[][][][] nt_fvs,
					       double[][][][] nt_probs,
					       Parameters params) throws IOException {
	int length = inst.length;
		
	// Get production crap.		
	for(int w1 = 0; w1 < length; w1++) {
	    for(int w2 = w1+1; w2 < length; w2++) {
				
		for(int ph = 0; ph < 2; ph++) {

		    FeatureVector prodFV = new FeatureVector(-1,-1.0,null);
					
		    int indx = in.readInt();
		    while(indx != -2) {
			prodFV = new FeatureVector(indx,1.0,prodFV);
			indx = in.readInt();
		    }
					
		    double prodProb = params.getScore(prodFV);
		    fvs[w1][w2][ph] = prodFV;
		    probs[w1][w2][ph] = prodProb;
		}
	    }
			
	}
	int last = in.readInt();
	if(last != -3) { System.out.println("Error reading file."); System.exit(0); }

	if(labeled) {
	    for(int w1 = 0; w1 < length; w1++) {
		
		for(int t = 0; t < types.length; t++) {
		    String type = types[t];
		    
		    for(int ph = 0; ph < 2; ph++) {						
			
			for(int ch = 0; ch < 2; ch++) {						
			    
			    FeatureVector prodFV = new FeatureVector(-1,-1.0,null);
			    
			    int indx = in.readInt();
			    while(indx != -2) {
				prodFV = new FeatureVector(indx,1.0,prodFV);
				indx = in.readInt();
			    }
			    
			    double nt_prob = params.getScore(prodFV);
			    nt_fvs[w1][t][ph][ch] = prodFV;
			    nt_probs[w1][t][ph][ch] = nt_prob;
			    
			}
		    }
		}
		
	    }
	    last = in.readInt();
	    if(last != -3) { System.out.println("Error reading file."); System.exit(0); }
	}

	FeatureVector nfv = new FeatureVector(-1,-1.0,null);
	int next = in.readInt();
	while(next != -4) {
	    nfv = new FeatureVector(next,1.0,nfv);
	    next = in.readInt();
	}

	String[] toks = null;
	String[] pos = null;
	String[] labs = null;
	String actParseTree = null;
	try {
	    toks = (String[])in.readObject();
	    next = in.readInt();
	    pos = (String[])in.readObject();
	    next = in.readInt();
	    labs = (String[])in.readObject();
	    next = in.readInt();
	    actParseTree = (String)in.readObject();
	    next = in.readInt();
	}
	catch(ClassNotFoundException e) { System.out.println("Error reading file."); System.exit(0); }
		
	if(next != -1) { System.out.println("Error reading file."); System.exit(0); }

	DependencyInstance pti = new DependencyInstance(toks,pos,labs,nfv);
	pti.actParseTree = actParseTree;
	return pti;
		
    }
		
    public void getFeatureVector(DependencyInstance inst,
				 FeatureVector[][][] fvs,
				 double[][][] probs,
				 FeatureVector[][][][] nt_fvs,
				 double[][][][] nt_probs, Parameters params) {

	String[] toks = inst.sentence;
	String[] pos = inst.pos;
	String[] labs = inst.labs;
		
	String[] posA = new String[pos.length];
	for(int i = 0; i < pos.length; i++) {
	    posA[i] = pos[i].substring(0,1);
	}

	// Get production crap.		
	for(int w1 = 0; w1 < toks.length; w1++) {
	    for(int w2 = w1+1; w2 < toks.length; w2++) {
				
		for(int ph = 0; ph < 2; ph++) {
		    boolean attR = ph == 0 ? true : false;
		    
		    int childInt = attR ? w2 : w1;
		    int parInt = attR ? w1 : w2;
		    
		    FeatureVector prodFV = createFeatureVector(toks,pos,posA,w1,w2,attR,
								    new FeatureVector(-1,-1.0,null));
										
		    double prodProb = params.getScore(prodFV);
		    fvs[w1][w2][ph] = prodFV;
		    probs[w1][w2][ph] = prodProb;
		}
	    }
			
	}

	if(labeled) {
	    for(int w1 = 0; w1 < toks.length; w1++) {
		
		for(int t = 0; t < types.length; t++) {
		    String type = types[t];
		    
		    for(int ph = 0; ph < 2; ph++) {						
			boolean attR = ph == 0 ? true : false;
			
			for(int ch = 0; ch < 2; ch++) {						
			    boolean child = ch == 0 ? true : false;						
			    
			    FeatureVector prodFV = createFeatureVector(toks,pos,posA,w1,
									    type,attR,child,
									    new FeatureVector(-1,-1.0,null));
			    
			    double nt_prob = params.getScore(prodFV);
			    nt_fvs[w1][t][ph][ch] = prodFV;
			    nt_probs[w1][t][ph][ch] = nt_prob;
			    
			}
		    }
		}
		
	    }
	}		
    }
		
}