Package opennlp.tools.parser

Examples of opennlp.tools.parser.Parse

61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
      if (line.equals("")) {
        System.out.println();
        clearPrevTokenMaps(finders);
        continue;
      }
      Parse p = Parse.parseParse(line);
      Parse[] tagNodes = p.getTagNodes();
      String[] tokens = new String[tagNodes.length];
      for (int ti=0;ti<tagNodes.length;ti++){
        tokens[ti] = tagNodes[ti].getCoveredText();
      }
      //System.err.println(java.util.Arrays.asList(tokens));
      for (int fi = 0, fl = finders.length; fi < fl; fi++) {
        nameSpans[fi] = finders[fi].nameFinder.find(tokens);
        //System.err.println("english.NameFinder.processParse: "+tags[fi] + " " + java.util.Arrays.asList(nameSpans[fi]));
      }
     
      for (int fi = 0, fl = finders.length; fi < fl; fi++) {
        Parse.addNames(tags[fi],nameSpans[fi],tagNodes);
      }
      p.show();
    }
  }
View Full Code Here
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
      return new DefaultParse(p,sentenceNumber);
    }
  }

  public opennlp.tools.coref.mention.Parse getNextToken() {
    Parse parent = parse.getParent();
    Parse node = parse;
    int index=-1;
    //find parent with subsequent children
    while(parent != null) {
      index = parent.indexOf(node)+1;
      if (index == parent.getChildCount()) {
        node = parent;
        parent = parent.getParent();
      }
      else {
        break;
      }
    }
    //find left-most child which is a token
    if (parent == null) {
      return null;
    }
    else {
      Parse p = parent.getChildren()[index];
      while (!p.isPosTag()) {
        p = p.getChildren()[0];
      }
      return new DefaultParse(p,sentenceNumber);
    }
  }
View Full Code Here
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
  @Override
  protected Parse[] advanceParses(final Parse p, double probMass) {
    double q = 1 - probMass;
    /** The closest previous node which has been labeled as a start node. */
    Parse lastStartNode = null;
    /** The index of the closest previous node which has been labeled as a start node. */
    int lastStartIndex = -1;
    /** The type of the closest previous node which has been labeled as a start node. */
    String lastStartType = null;
    /** The index of the node which will be labeled in this iteration of advancing the parse. */
    int advanceNodeIndex;
    /** The node which will be labeled in this iteration of advancing the parse. */
    Parse advanceNode=null;
    Parse[] originalChildren = p.getChildren();
    Parse[] children = collapsePunctuation(originalChildren,punctSet);
    int numNodes = children.length;
    if (numNodes == 0) {
      return null;
    }
    //determines which node needs to be labeled and prior labels.
    for (advanceNodeIndex = 0; advanceNodeIndex < numNodes; advanceNodeIndex++) {
      advanceNode = children[advanceNodeIndex];
      if (advanceNode.getLabel() == null) {
        break;
      }
      else if (startTypeMap.containsKey(advanceNode.getLabel())) {
        lastStartType = startTypeMap.get(advanceNode.getLabel());
        lastStartNode = advanceNode;
        lastStartIndex = advanceNodeIndex;
        //System.err.println("lastStart "+i+" "+lastStart.label+" "+lastStart.prob);
      }
    }
    int originalAdvanceIndex = mapParseIndex(advanceNodeIndex,children,originalChildren);
    List<Parse> newParsesList = new ArrayList<Parse>(buildModel.getNumOutcomes());
    //call build
    buildModel.eval(buildContextGenerator.getContext(children, advanceNodeIndex), bprobs);
    double bprobSum = 0;
    while (bprobSum < probMass) {
      // The largest unadvanced labeling.
      int max = 0;
      for (int pi = 1; pi < bprobs.length; pi++) { //for each build outcome
        if (bprobs[pi] > bprobs[max]) {
          max = pi;
        }
      }
      if (bprobs[max] == 0) {
        break;
      }
      double bprob = bprobs[max];
      bprobs[max] = 0; //zero out so new max can be found
      bprobSum += bprob;
      String tag = buildModel.getOutcome(max);
      //System.out.println("trying "+tag+" "+bprobSum+" lst="+lst);
      if (max == topStartIndex) { // can't have top until complete
        continue;
      }
      //System.err.println(i+" "+tag+" "+bprob);
      if (startTypeMap.containsKey(tag)) { //update last start
        lastStartIndex = advanceNodeIndex;
        lastStartNode = advanceNode;
        lastStartType = startTypeMap.get(tag);
      }
      else if (contTypeMap.containsKey(tag)) {
        if (lastStartNode == null || !lastStartType.equals(contTypeMap.get(tag))) {
          continue; //Cont must match previous start or continue
        }
      }
      Parse newParse1 = (Parse) p.clone(); //clone parse
      if (createDerivationString) newParse1.getDerivation().append(max).append("-");
      newParse1.setChild(originalAdvanceIndex,tag); //replace constituent being labeled to create new derivation
      newParse1.addProb(Math.log(bprob));
      //check
      //String[] context = checkContextGenerator.getContext(newParse1.getChildren(), lastStartType, lastStartIndex, advanceNodeIndex);
      checkModel.eval(checkContextGenerator.getContext(collapsePunctuation(newParse1.getChildren(),punctSet), lastStartType, lastStartIndex, advanceNodeIndex), cprobs);
      //System.out.println("check "+lastStartType+" "+cprobs[completeIndex]+" "+cprobs[incompleteIndex]+" "+tag+" "+java.util.Arrays.asList(context));
      Parse newParse2 = newParse1;
      if (cprobs[completeIndex] > q) { //make sure a reduce is likely
        newParse2 = (Parse) newParse1.clone();
        if (createDerivationString) newParse2.getDerivation().append(1).append(".");
        newParse2.addProb(Math.log(cprobs[completeIndex]));
        Parse[] cons = new Parse[advanceNodeIndex - lastStartIndex + 1];
        boolean flat = true;
        //first
        cons[0] = lastStartNode;
        flat &= cons[0].isPosTag();
        //last
        cons[advanceNodeIndex - lastStartIndex] = advanceNode;
        flat &= cons[advanceNodeIndex - lastStartIndex].isPosTag();
        //middle
        for (int ci = 1; ci < advanceNodeIndex - lastStartIndex; ci++) {
          cons[ci] = children[ci + lastStartIndex];
          flat &= cons[ci].isPosTag();
        }
        if (!flat) { //flat chunks are done by chunker
          if (lastStartIndex == 0 && advanceNodeIndex == numNodes-1) { //check for top node to include end and begining punctuation
            //System.err.println("ParserME.advanceParses: reducing entire span: "+new Span(lastStartNode.getSpan().getStart(), advanceNode.getSpan().getEnd())+" "+lastStartType+" "+java.util.Arrays.asList(children));
            newParse2.insert(new Parse(p.getText(), p.getSpan(), lastStartType, cprobs[1], headRules.getHead(cons, lastStartType)));
          }
          else {
            newParse2.insert(new Parse(p.getText(), new Span(lastStartNode.getSpan().getStart(), advanceNode.getSpan().getEnd()), lastStartType, cprobs[1], headRules.getHead(cons, lastStartType)));
          }
          newParsesList.add(newParse2);
        }
      }
      if (cprobs[incompleteIndex] > q) { //make sure a shift is likely
View Full Code Here
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
    //default
    features.add("default");
    //first constituent label
    features.add("fl="+constituents[0].getLabel());
    Parse pstart = constituents[start];
    Parse pend = constituents[end];
    checkcons(pstart, "begin", type, features);
    checkcons(pend, "last", type, features);
    StringBuilder production = new StringBuilder(20);
    StringBuilder punctProduction = new StringBuilder(20);
    production.append("p=").append(type).append("->");
    punctProduction.append("pp=").append(type).append("->");
    for (int pi = start; pi < end; pi++) {
      Parse p = constituents[pi];
      checkcons(p, pend, type, features);
      production.append(p.getType()).append(",");
      punctProduction.append(p.getType()).append(",");
      Collection<Parse> nextPunct = p.getNextPunctuationSet();
      if (nextPunct != null) {
        for (Iterator<Parse> pit=nextPunct.iterator();pit.hasNext();) {
          Parse punct = pit.next();
          punctProduction.append(punct.getType()).append(",");
        }
      }
    }
    production.append(pend.getType());
    punctProduction.append(pend.getType());
    features.add(production.toString());
    features.add(punctProduction.toString());
    Parse p_2 = null;
    Parse p_1 = null;
    Parse p1 = null;
    Parse p2 = null;
    Collection<Parse> p1s = constituents[end].getNextPunctuationSet();
    Collection<Parse> p2s = null;
    Collection<Parse> p_1s = constituents[start].getPreviousPunctuationSet();
    Collection<Parse> p_2s = null;
    if (start - 2 >= 0) {
View Full Code Here
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
    super(samples);
  }

  public POSSample read() throws IOException {
   
    Parse parse = samples.read();
   
    if (parse != null) {
     
      List<String> sentence = new ArrayList<String>();
      List<String> tags = new ArrayList<String>();
     
      for(Parse tagNode : parse.getTagNodes()) {
        sentence.add(tagNode.getCoveredText());
        tags.add(tagNode.getType());
      }
     
      return new POSSample(sentence, tags);
View Full Code Here
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
    List<String> features = new ArrayList<String>(100);
    int ps = constituents.length;

    // cons(-2), cons(-1), cons(0), cons(1), cons(2)
    // cons(-2)
    Parse p_2 = null;
    Parse p_1 = null;
    Parse p0 = null;
    Parse p1 = null;
    Parse p2 = null;

    Collection<Parse> punct1s = null;
    Collection<Parse> punct2s = null;
    Collection<Parse> punct_1s = null;
    Collection<Parse> punct_2s = null;

    if (index - 2 >= 0) {
      p_2 = constituents[index - 2];
    }
    if (index - 1 >= 0) {
      p_1 = constituents[index - 1];
      punct_2s = p_1.getPreviousPunctuationSet();
    }
    p0 = constituents[index];
    punct_1s=p0.getPreviousPunctuationSet();
    punct1s=p0.getNextPunctuationSet();

    if (index + 1 < ps) {
      p1 = constituents[index + 1];
      punct2s = p1.getNextPunctuationSet();
    }
    if (index + 2 < ps) {
      p2 = constituents[index + 2];
    }

    boolean u_2 = true;
    boolean u_1 = true;
    boolean u0 = true;
    boolean u1 = true;
    boolean u2 = true;
    boolean b_2_1 = true;
    boolean b_10 = true;
    boolean b01 = true;
    boolean b12 = true;
    boolean t_2_10 = true;
    boolean t_101 = true;
    boolean t012 = true;

    if (dict != null) {

      if (p_2 != null) {
        unigram[0] = p_2.getHead().getCoveredText();
        u_2 = dict.contains(new StringList(unigram));
      }

      if (p2 != null) {
        unigram[0] = p2.getHead().getCoveredText();
        u2 = dict.contains(new StringList(unigram));
      }

      unigram[0] = p0.getHead().getCoveredText();
      u0 = dict.contains(new StringList(unigram));

      if (p_2 != null && p_1 != null) {
        bigram[0] = p_2.getHead().getCoveredText();
        bigram[1] = p_1.getHead().getCoveredText();
        b_2_1 = dict.contains(new StringList(bigram));

        trigram[0] = p_2.getHead().getCoveredText();
        trigram[1] = p_1.getHead().getCoveredText();
        trigram[2] = p0.getHead().getCoveredText();
        t_2_10 = dict.contains(new StringList(trigram));
      }
      if (p_1 != null && p1 != null) {
        trigram[0] = p_1.getHead().getCoveredText();
        trigram[1] = p0.getHead().getCoveredText();
        trigram[2] = p1.getHead().getCoveredText();
        t_101 = dict.contains(new StringList(trigram));
      }
      if (p_1 != null) {
        unigram[0] = p_1.getHead().getCoveredText();
        u_1 = dict.contains(new StringList(unigram));

        //extra check for 2==null case
        b_2_1 = b_2_1 && u_1 & u_2;
        t_2_10 = t_2_10 && u_1 & u_2 & u0;
        t_101 = t_101 && u_1 & u0 && u1;

        bigram[0] = p_1.getHead().getCoveredText();
        bigram[1] = p0.getHead().getCoveredText();
        b_10 = dict.contains(new StringList(bigram)) && u_1 && u0;
      }
      if (p1 != null && p2 != null) {
        bigram[0] = p1.getHead().getCoveredText();
        bigram[1] = p2.getHead().getCoveredText();
        b12 = dict.contains(new StringList(bigram));

        trigram[0] = p0.getHead().getCoveredText();
        trigram[1] = p1.getHead().getCoveredText();
        trigram[2] = p2.getHead().getCoveredText();
        t012 = dict.contains(new StringList(trigram));
      }
      if (p1 != null) {
        unigram[0] = p1.getHead().getCoveredText();
        u1 = dict.contains(new StringList(unigram));

        //extra check for 2==null case
        b12 = b12 && u1 && u2;
        t012 = t012 && u1 && u2 && u0;
        t_101 = t_101 && u0 && u_1 && u1;

        bigram[0] = p0.getHead().getCoveredText();
        bigram[1] = p1.getHead().getCoveredText();
        b01 = dict.contains(new StringList(bigram));
        b01 = b01 && u0 && u1;
      }
    }

    String consp_2 = cons(p_2, -2);
    String consp_1 = cons(p_1, -1);
    String consp0 = cons(p0, 0);
    String consp1 = cons(p1, 1);
    String consp2 = cons(p2, 2);

    String consbop_2 = consbo(p_2, -2);
    String consbop_1 = consbo(p_1, -1);
    String consbop0 = consbo(p0, 0);
    String consbop1 = consbo(p1, 1);
    String consbop2 = consbo(p2, 2);

    Cons c_2 = new Cons(consp_2,consbop_2,-2,u_2);
    Cons c_1 = new Cons(consp_1,consbop_1,-1,u_1);
    Cons c0 = new Cons(consp0,consbop0,0,u0);
    Cons c1 = new Cons(consp1,consbop1,1,u1);
    Cons c2 = new Cons(consp2,consbop2,2,u2);

    //default
    features.add("default");
    //first constituent label
    //features.add("fl="+constituents[0].getLabel());

    // features.add("stage=cons(i)");
    // cons(-2), cons(-1), cons(0), cons(1), cons(2)
    if (u0) features.add(consp0);
    features.add(consbop0);

    if (u_2) features.add(consp_2);
    features.add(consbop_2);
    if (u_1) features.add(consp_1);
    features.add(consbop_1);
    if (u1) features.add(consp1);
    features.add(consbop1);
    if (u2) features.add(consp2);
    features.add(consbop2);

    //cons(0),cons(1)
    cons2(features,c0,c1,punct1s,b01);
    //cons(-1),cons(0)
    cons2(features,c_1,c0,punct_1s,b_10);
    //features.add("stage=cons(0),cons(1),cons(2)");
    cons3(features,c0,c1,c2,punct1s,punct2s,t012,b01,b12);
    cons3(features,c_2,c_1,c0,punct_2s,punct_1s,t_2_10,b_2_1,b_10);
    cons3(features,c_1,c0,c1,punct_1s,punct1s,t_101,b_10,b01);
    //features.add("stage=other");
    String p0Tag = p0.getType();
    if (p0Tag.equals("-RRB-")) {
      for (int pi = index - 1; pi >= 0; pi--) {
        Parse p = constituents[pi];
        if (p.getType().equals("-LRB-")) {
          features.add("bracketsmatch");
          break;
        }
        if (p.getLabel().startsWith(Parser.START)) {
          break;
        }
      }
    }
    if (p0Tag.equals("-RCB-")) {
      for (int pi = index - 1; pi >= 0; pi--) {
        Parse p = constituents[pi];
        if (p.getType().equals("-LCB-")) {
          features.add("bracketsmatch");
          break;
        }
        if (p.getLabel().startsWith(Parser.START)) {
          break;
        }
      }
    }
    if (p0Tag.equals("''")) {
      for (int pi = index - 1; pi >= 0; pi--) {
        Parse p = constituents[pi];
        if (p.getType().equals("``")) {
          features.add("quotesmatch");
          break;
        }
        if (p.getLabel().startsWith(Parser.START)) {
          break;
        }
      }
    }
    if (p0Tag.equals("'")) {
      for (int pi = index - 1; pi >= 0; pi--) {
        Parse p = constituents[pi];
        if (p.getType().equals("`")) {
          features.add("quotesmatch");
          break;
        }
        if (p.getLabel().startsWith(Parser.START)) {
          break;
        }
      }
    }
    if (p0Tag.equals(",")) {
      for (int pi = index - 1; pi >= 0; pi--) {
        Parse p = constituents[pi];
        if (p.getType().equals(",")) {
          features.add("iscomma");
          break;
        }
        if (p.getLabel().startsWith(Parser.START)) {
          break;
        }
      }
    }
    if (p0Tag.equals(".") && index == ps - 1) {
      for (int pi = index - 1; pi >= 0; pi--) {
        Parse p = constituents[pi];
        if (p.getLabel().startsWith(Parser.START)) {
          if (pi == 0) {
            features.add("endofsentence");
          }
          break;
        }
View Full Code Here
54
55
56
57
58
59
60
61
62
63
64
     
      List<Parse> parses = new ArrayList<Parse>();
     
      for (Parse p : sample.getParses()) {
       
        Parse parseTokens[] = p.getTagNodes();
        String tokens[] = new String[parseTokens.length];
       
        for (int i = 0; i < tokens.length; i++) {
          tokens[i] = parseTokens[i].toString();
        }
View Full Code Here
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
   * @param root The root of the parse tree.
   * @return The right frontier of the specified parse tree.
   */
  public static List<Parse> getRightFrontier(Parse root,Set<String> punctSet) {
    List<Parse> rf = new LinkedList<Parse>();
    Parse top;
    if (root.getType() == AbstractBottomUpParser.TOP_NODE ||
        root.getType() == AbstractBottomUpParser.INC_NODE) {
      top = collapsePunctuation(root.getChildren(),punctSet)[0];
    }
    else {
      top = root;
    }
    while(!top.isPosTag()) {
      rf.add(0,top);
      Parse[] kids = top.getChildren();
      top = kids[kids.length-1];
    }
    return new ArrayList<Parse>(rf);
  }
View Full Code Here
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
  protected Parse[] advanceParses(Parse p, double probMass) {
    double q = 1 - probMass;
    /** The index of the node which will be labeled in this iteration of advancing the parse. */
    int advanceNodeIndex;
    /** The node which will be labeled in this iteration of advancing the parse. */
    Parse advanceNode=null;
    Parse[] originalChildren = p.getChildren();
    Parse[] children = collapsePunctuation(originalChildren,punctSet);
    int numNodes = children.length;
    if (numNodes == 0) {
      return null;
    }
    else if (numNodes == 1) {  //put sentence initial and final punct in top node
      if (children[0].isPosTag()) {
        return null;
      }
      else {
        p.expandTopNode(children[0]);
        return new Parse[] { p };
      }
    }
    //determines which node needs to adanced.
    for (advanceNodeIndex = 0; advanceNodeIndex < numNodes; advanceNodeIndex++) {
      advanceNode = children[advanceNodeIndex];
      if (!isBuilt(advanceNode)) {
        break;
      }
    }
    int originalZeroIndex = mapParseIndex(0,children,originalChildren);
    int originalAdvanceIndex = mapParseIndex(advanceNodeIndex,children,originalChildren);
    List<Parse> newParsesList = new ArrayList<Parse>();
    //call build model
    buildModel.eval(buildContextGenerator.getContext(children, advanceNodeIndex), bprobs);
    double doneProb = bprobs[doneIndex];
    if (debugOn) System.out.println("adi="+advanceNodeIndex+" "+advanceNode.getType()+"."+advanceNode.getLabel()+" "+advanceNode+" choose build="+(1-doneProb)+" attach="+doneProb);
    if (1-doneProb > q) {
      double bprobSum = 0;
      while (bprobSum < probMass) {
        /** The largest unadvanced labeling. */
        int max = 0;
        for (int pi = 1; pi < bprobs.length; pi++) { //for each build outcome
          if (bprobs[pi] > bprobs[max]) {
            max = pi;
          }
        }
        if (bprobs[max] == 0) {
          break;
        }
        double bprob = bprobs[max];
        bprobs[max] = 0; //zero out so new max can be found
        bprobSum += bprob;
        String tag = buildModel.getOutcome(max);
        if (!tag.equals(DONE)) {
          Parse newParse1 = (Parse) p.clone();
          Parse newNode = new Parse(p.getText(),advanceNode.getSpan(),tag,bprob,advanceNode.getHead());
          newParse1.insert(newNode);
          newParse1.addProb(Math.log(bprob));
          newParsesList.add(newParse1);
          if (checkComplete) {
            cprobs = checkModel.eval(checkContextGenerator.getContext(newNode,children,advanceNodeIndex,false));
            if (debugOn) System.out.println("building "+tag+" "+bprob+" c="+cprobs[completeIndex]);
            if (cprobs[completeIndex] > probMass) { //just complete advances
              setComplete(newNode);
              newParse1.addProb(Math.log(cprobs[completeIndex]));
              if (debugOn) System.out.println("Only advancing complete node");
            }
            else if (1-cprobs[completeIndex] > probMass) { //just incomplete advances
              setIncomplete(newNode);
              newParse1.addProb(Math.log(1-cprobs[completeIndex]));
              if (debugOn) System.out.println("Only advancing incomplete node");
            }
            else { //both complete and incomplete advance
              if (debugOn) System.out.println("Advancing both complete and incomplete nodes");
              setComplete(newNode);
              newParse1.addProb(Math.log(cprobs[completeIndex]));

              Parse newParse2 = (Parse) p.clone();
              Parse newNode2 = new Parse(p.getText(),advanceNode.getSpan(),tag,bprob,advanceNode.getHead());
              newParse2.insert(newNode2);
              newParse2.addProb(Math.log(bprob));
              newParsesList.add(newParse2);
              newParse2.addProb(Math.log(1-cprobs[completeIndex]));
              setIncomplete(newNode2); //set incomplete for non-clone
            }
          }
          else {
            if (debugOn) System.out.println("building "+tag+" "+bprob);
          }
        }
      }
    }
    //advance attaches
    if (doneProb > q) {
      Parse newParse1 = (Parse) p.clone(); //clone parse
      //mark nodes as built
      if (checkComplete) {
        if (isComplete(advanceNode)) {
          newParse1.setChild(originalAdvanceIndex,Parser.BUILT+"."+Parser.COMPLETE); //replace constituent being labeled to create new derivation
        }
        else {
          newParse1.setChild(originalAdvanceIndex,Parser.BUILT+"."+Parser.INCOMPLETE); //replace constituent being labeled to create new derivation
        }
      }
      else {
        newParse1.setChild(originalAdvanceIndex,Parser.BUILT); //replace constituent being labeled to create new derivation
      }
      newParse1.addProb(Math.log(doneProb));
      if (advanceNodeIndex == 0) { //no attach if first node.
        newParsesList.add(newParse1);
      }
      else {
        List<Parse> rf = getRightFrontier(p,punctSet);
        for (int fi=0,fs=rf.size();fi<fs;fi++) {
          Parse fn = rf.get(fi);
          attachModel.eval(attachContextGenerator.getContext(children, advanceNodeIndex,rf,fi), aprobs);
          if (debugOn) {
            //List cs = java.util.Arrays.asList(attachContextGenerator.getContext(children, advanceNodeIndex,rf,fi,punctSet));
            System.out.println("Frontier node("+fi+"): "+fn.getType()+"."+fn.getLabel()+" "+fn+" <- "+advanceNode.getType()+" "+advanceNode+" d="+aprobs[daughterAttachIndex]+" s="+aprobs[sisterAttachIndex]+" ");
          }
          for (int ai=0;ai<attachments.length;ai++) {
            double prob = aprobs[attachments[ai]];
            //should we try an attach if p > threshold and
            // if !checkComplete then prevent daughter attaching to chunk
            // if checkComplete then prevent daughter attacing to complete node or
            //    sister attaching to an incomplete node
            if (prob > q && (
                (!checkComplete && (attachments[ai]!= daughterAttachIndex || !isComplete(fn)))
                ||
                (checkComplete && ((attachments[ai]== daughterAttachIndex && !isComplete(fn)) || (attachments[ai] == sisterAttachIndex && isComplete(fn)))))) {
              Parse newParse2 = newParse1.cloneRoot(fn,originalZeroIndex);
              Parse[] newKids = Parser.collapsePunctuation(newParse2.getChildren(),punctSet);
              //remove node from top level since were going to attach it (including punct)
              for (int ri=originalZeroIndex+1;ri<=originalAdvanceIndex;ri++) {
                //System.out.println(at"-removing "+(originalZeroIndex+1)+" "+newParse2.getChildren()[originalZeroIndex+1]);
                newParse2.remove(originalZeroIndex+1);
              }
              List<Parse> crf = getRightFrontier(newParse2,punctSet);
              Parse updatedNode;
              if (attachments[ai] == daughterAttachIndex) {//attach daughter
                updatedNode = crf.get(fi);
                updatedNode.add(advanceNode,headRules);
              }
              else { //attach sister
                Parse psite;
                if (fi+1 < crf.size()) {
                  psite = crf.get(fi+1);
                  updatedNode = psite.adjoin(advanceNode,headRules);
                }
                else {
                  psite = newParse2;
                  updatedNode = psite.adjoinRoot(advanceNode,headRules,originalZeroIndex);
                  newKids[0] = updatedNode;
                }
              }
              //update spans affected by attachment
              for (int ni=fi+1;ni<crf.size();ni++) {
                Parse node = crf.get(ni);
                node.updateSpan();
              }
              //if (debugOn) {System.out.print(ai+"-result: ");newParse2.show();System.out.println();}
              newParse2.addProb(Math.log(prob));
              newParsesList.add(newParse2);
              if (checkComplete) {
                cprobs = checkModel.eval(checkContextGenerator.getContext(updatedNode,newKids,advanceNodeIndex,true));
                if (cprobs[completeIndex] > probMass) {
                  setComplete(updatedNode);
                  newParse2.addProb(Math.log(cprobs[completeIndex]));
                  if (debugOn) System.out.println("Only advancing complete node");
                }
                else if (1-cprobs[completeIndex] > probMass) {
                  setIncomplete(updatedNode);
                  newParse2.addProb(Math.log(1-cprobs[completeIndex]));
                  if (debugOn) System.out.println("Only advancing incomplete node");
                }
                else {
                  setComplete(updatedNode);
                  Parse newParse3 = newParse2.cloneRoot(updatedNode,originalZeroIndex);
                  newParse3.addProb(Math.log(cprobs[completeIndex]));
                  newParsesList.add(newParse3);
                  setIncomplete(updatedNode);
                  newParse2.addProb(Math.log(1-cprobs[completeIndex]));
                  if (debugOn) System.out.println("Advancing both complete and incomplete nodes; c="+cprobs[completeIndex]);
                }
View Full Code Here
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
   
    List<Mention> mentions = new ArrayList<Mention>();
   
    for (opennlp.tools.coref.mention.Parse corefParse : sample.getParses()) {

      Parse p = ((DefaultParse) corefParse).getParse();
     
      Mention extents[] = mentionFinder.getMentions(corefParse);
     
      for (int ei = 0, en = extents.length; ei < en;ei++) {

        if (extents[ei].getParse() == null) {

          Stack<Parse> nodes = new Stack<Parse>();
          nodes.add(p);
         
          while (!nodes.isEmpty()) {
           
            Parse node = nodes.pop();
           
            if (node.getSpan().equals(extents[ei].getSpan()) && node.getType().startsWith("NML")) {
              DefaultParse corefParseNode = new DefaultParse(node, corefParse.getSentenceNumber());
              extents[ei].setParse(corefParseNode);
              extents[ei].setId(corefParseNode.getEntityId());
              break;
            }
           
            nodes.addAll(Arrays.asList(node.getChildren()));
          }
        }
      }
     
      mentions.addAll(Arrays.asList(extents));
View Full Code Here
TOP

Related Classes of opennlp.tools.parser.Parse

  • com.tamingtext.opennlp.ChunkParserTest
  • com.tamingtext.qa.AnswerTypeContextGenerator
  • com.tamingtext.qa.AnswerTypeEventStream
  • com.tamingtext.qa.ChunkParser
  • com.tamingtext.qa.QuestionQParser
  • functionality.Parsing
  • net.sf.nlpshell.CorefParse
  • net.sf.nlpshell.Main
  • opennlp.tools.cmdline.coref.CoreferencerTool
  • opennlp.tools.cmdline.coref.CoreferencerTool$CorefParse
    • Copyright © 2018 www.massapicom. All rights reserved.
    All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.