1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
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
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
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
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
|
/*
* extract.cpp
* Modified by: Rohit Gupta CDAC, Mumbai, India
* on July 15, 2012 to implement parallel processing
* Modified by: Nadi Tomeh - LIMSI/CNRS
* Machine Translation Marathon 2010, Dublin
*/
#include <cstdio>
#include <iostream>
#include <fstream>
#include <vector>
#include <string>
#include <stdlib.h>
#include <assert.h>
#include <cstring>
#include <sstream>
#include <map>
#include <set>
#include <vector>
#include "SafeGetline.h"
#include "SentenceAlignment.h"
#include "tables-core.h"
#include "InputFileStream.h"
#include "OutputFileStream.h"
#include "PhraseExtractionOptions.h"
using namespace std;
using namespace MosesTraining;
namespace MosesTraining {
const long int LINE_MAX_LENGTH = 500000 ;
// HPhraseVertex represents a point in the alignment matrix
typedef pair <int, int> HPhraseVertex;
// Phrase represents a bi-phrase; each bi-phrase is defined by two points in the alignment matrix:
// bottom-left and top-right
typedef pair<HPhraseVertex, HPhraseVertex> HPhrase;
// HPhraseVector is a vector of HPhrases
typedef vector < HPhrase > HPhraseVector;
// SentenceVertices represents, from all extracted phrases, all vertices that have the same positioning
// The key of the map is the English index and the value is a set of the source ones
typedef map <int, set<int> > HSentenceVertices;
REO_POS getOrientWordModel(SentenceAlignment &, REO_MODEL_TYPE, bool, bool,
int, int, int, int, int, int, int,
bool (*)(int, int), bool (*)(int, int));
REO_POS getOrientPhraseModel(SentenceAlignment &, REO_MODEL_TYPE, bool, bool,
int, int, int, int, int, int, int,
bool (*)(int, int), bool (*)(int, int),
const HSentenceVertices &, const HSentenceVertices &);
REO_POS getOrientHierModel(SentenceAlignment &, REO_MODEL_TYPE, bool, bool,
int, int, int, int, int, int, int,
bool (*)(int, int), bool (*)(int, int),
const HSentenceVertices &, const HSentenceVertices &,
const HSentenceVertices &, const HSentenceVertices &,
REO_POS);
void insertVertex(HSentenceVertices &, int, int);
void insertPhraseVertices(HSentenceVertices &, HSentenceVertices &, HSentenceVertices &, HSentenceVertices &,
int, int, int, int);
string getOrientString(REO_POS, REO_MODEL_TYPE);
bool ge(int, int);
bool le(int, int);
bool lt(int, int);
bool isAligned (SentenceAlignment &, int, int);
int sentenceOffset = 0;
}
namespace MosesTraining{
class ExtractTask
{
public:
ExtractTask(size_t id, SentenceAlignment &sentence,PhraseExtractionOptions &initoptions, Moses::OutputFileStream &extractFile, Moses::OutputFileStream &extractFileInv,Moses::OutputFileStream &extractFileOrientation):
m_sentence(sentence),
m_options(initoptions),
m_extractFile(extractFile),
m_extractFileInv(extractFileInv),
m_extractFileOrientation(extractFileOrientation){}
void Run();
private:
vector< string > m_extractedPhrases;
vector< string > m_extractedPhrasesInv;
vector< string > m_extractedPhrasesOri;
vector< string > m_extractedPhrasesSid;
void extractBase(SentenceAlignment &);
void extract(SentenceAlignment &);
void addPhrase(SentenceAlignment &, int, int, int, int, string &);
void writePhrasesToFile();
SentenceAlignment &m_sentence;
const PhraseExtractionOptions &m_options;
Moses::OutputFileStream &m_extractFile;
Moses::OutputFileStream &m_extractFileInv;
Moses::OutputFileStream &m_extractFileOrientation;
};
}
int main(int argc, char* argv[])
{
cerr << "PhraseExtract v1.4, written by Philipp Koehn\n"
<< "phrase extraction from an aligned parallel corpus\n";
if (argc < 6) {
cerr << "syntax: extract en de align extract max-length [orientation [ --model [wbe|phrase|hier]-[msd|mslr|mono] ] ";
cerr<<"| --OnlyOutputSpanInfo | --NoTTable | --GZOutput | --IncludeSentenceId | --SentenceOffset n | --InstanceWeights filename ]\n";
exit(1);
}
Moses::OutputFileStream extractFile;
Moses::OutputFileStream extractFileInv;
Moses::OutputFileStream extractFileOrientation;
const char* const &fileNameE = argv[1];
const char* const &fileNameF = argv[2];
const char* const &fileNameA = argv[3];
const string fileNameExtract = string(argv[4]);
PhraseExtractionOptions options(atoi(argv[5]));
for(int i=6; i<argc; i++) {
if (strcmp(argv[i],"--OnlyOutputSpanInfo") == 0) {
options.initOnlyOutputSpanInfo(true);
} else if (strcmp(argv[i],"orientation") == 0 || strcmp(argv[i],"--Orientation") == 0) {
options.initOrientationFlag(true);
} else if (strcmp(argv[i],"--NoTTable") == 0) {
options.initTranslationFlag(false);
} else if (strcmp(argv[i], "--IncludeSentenceId") == 0) {
options.initIncludeSentenceIdFlag(true);
} else if (strcmp(argv[i], "--SentenceOffset") == 0) {
if (i+1 >= argc || argv[i+1][0] < '0' || argv[i+1][0] > '9') {
cerr << "extract: syntax error, used switch --SentenceOffset without a number" << endl;
exit(1);
}
sentenceOffset = atoi(argv[++i]);
} else if (strcmp(argv[i], "--GZOutput") == 0) {
options.initGzOutput(true);
} else if (strcmp(argv[i], "--InstanceWeights") == 0) {
if (i+1 >= argc) {
cerr << "extract: syntax error, used switch --InstanceWeights without file name" << endl;
exit(1);
}
options.initInstanceWeightsFile(argv[++i]);
} else if(strcmp(argv[i],"--model") == 0) {
if (i+1 >= argc) {
cerr << "extract: syntax error, no model's information provided to the option --model " << endl;
exit(1);
}
char* modelParams = argv[++i];
char* modelName = strtok(modelParams, "-");
char* modelType = strtok(NULL, "-");
// REO_MODEL_TYPE intModelType;
if(strcmp(modelName, "wbe") == 0) {
options.initWordModel(true);
if(strcmp(modelType, "msd") == 0)
options.initWordType(REO_MSD);
else if(strcmp(modelType, "mslr") == 0)
options.initWordType(REO_MSLR);
else if(strcmp(modelType, "mono") == 0 || strcmp(modelType, "monotonicity") == 0)
options.initWordType(REO_MONO);
else {
cerr << "extract: syntax error, unknown reordering model type: " << modelType << endl;
exit(1);
}
} else if(strcmp(modelName, "phrase") == 0) {
options.initPhraseModel(true);
if(strcmp(modelType, "msd") == 0)
options.initPhraseType(REO_MSD);
else if(strcmp(modelType, "mslr") == 0)
options.initPhraseType(REO_MSLR);
else if(strcmp(modelType, "mono") == 0 || strcmp(modelType, "monotonicity") == 0)
options.initPhraseType(REO_MONO);
else {
cerr << "extract: syntax error, unknown reordering model type: " << modelType << endl;
exit(1);
}
} else if(strcmp(modelName, "hier") == 0) {
options.initHierModel(true);
if(strcmp(modelType, "msd") == 0)
options.initHierType(REO_MSD);
else if(strcmp(modelType, "mslr") == 0)
options.initHierType(REO_MSLR);
else if(strcmp(modelType, "mono") == 0 || strcmp(modelType, "monotonicity") == 0)
options.initHierType(REO_MONO);
else {
cerr << "extract: syntax error, unknown reordering model type: " << modelType << endl;
exit(1);
}
} else {
cerr << "extract: syntax error, unknown reordering model: " << modelName << endl;
exit(1);
}
options.initAllModelsOutputFlag(true);
} else {
cerr << "extract: syntax error, unknown option '" << string(argv[i]) << "'\n";
exit(1);
}
}
// default reordering model if no model selected
// allows for the old syntax to be used
if(options.isOrientationFlag() && !options.isAllModelsOutputFlag()) {
options.initWordModel(true);
options.initWordType(REO_MSD);
}
// open input files
Moses::InputFileStream eFile(fileNameE);
Moses::InputFileStream fFile(fileNameF);
Moses::InputFileStream aFile(fileNameA);
istream *eFileP = &eFile;
istream *fFileP = &fFile;
istream *aFileP = &aFile;
istream *iwFileP = NULL;
auto_ptr<Moses::InputFileStream> instanceWeightsFile;
if (options.getInstanceWeightsFile().length()) {
instanceWeightsFile.reset(new Moses::InputFileStream(options.getInstanceWeightsFile()));
iwFileP = instanceWeightsFile.get();
}
// open output files
if (options.isTranslationFlag()) {
string fileNameExtractInv = fileNameExtract + ".inv" + (options.isGzOutput()?".gz":"");
extractFile.Open( (fileNameExtract + (options.isGzOutput()?".gz":"")).c_str());
extractFileInv.Open(fileNameExtractInv.c_str());
}
if (options.isOrientationFlag()) {
string fileNameExtractOrientation = fileNameExtract + ".o" + (options.isGzOutput()?".gz":"");
extractFileOrientation.Open(fileNameExtractOrientation.c_str());
}
int i = sentenceOffset;
while(true) {
i++;
if (i%10000 == 0) cerr << "." << flush;
char englishString[LINE_MAX_LENGTH];
char foreignString[LINE_MAX_LENGTH];
char alignmentString[LINE_MAX_LENGTH];
char weightString[LINE_MAX_LENGTH];
SAFE_GETLINE((*eFileP), englishString, LINE_MAX_LENGTH, '\n', __FILE__);
if (eFileP->eof()) break;
SAFE_GETLINE((*fFileP), foreignString, LINE_MAX_LENGTH, '\n', __FILE__);
SAFE_GETLINE((*aFileP), alignmentString, LINE_MAX_LENGTH, '\n', __FILE__);
if (iwFileP) {
SAFE_GETLINE((*iwFileP), weightString, LINE_MAX_LENGTH, '\n', __FILE__);
}
SentenceAlignment sentence;
// cout << "read in: " << englishString << " & " << foreignString << " & " << alignmentString << endl;
//az: output src, tgt, and alingment line
if (options.isOnlyOutputSpanInfo()) {
cout << "LOG: SRC: " << foreignString << endl;
cout << "LOG: TGT: " << englishString << endl;
cout << "LOG: ALT: " << alignmentString << endl;
cout << "LOG: PHRASES_BEGIN:" << endl;
}
if (sentence.create( englishString, foreignString, alignmentString, weightString, i, false)) {
ExtractTask *task = new ExtractTask(i-1, sentence, options, extractFile , extractFileInv, extractFileOrientation);
task->Run();
delete task;
}
if (options.isOnlyOutputSpanInfo()) cout << "LOG: PHRASES_END:" << endl; //az: mark end of phrases
}
eFile.Close();
fFile.Close();
aFile.Close();
//az: only close if we actually opened it
if (!options.isOnlyOutputSpanInfo()) {
if (options.isTranslationFlag()) {
extractFile.Close();
extractFileInv.Close();
}
if (options.isOrientationFlag()){
extractFileOrientation.Close();
}
}
}
namespace MosesTraining
{
void ExtractTask::Run() {
extract(m_sentence);
writePhrasesToFile();
m_extractedPhrases.clear();
m_extractedPhrasesInv.clear();
m_extractedPhrasesOri.clear();
m_extractedPhrasesSid.clear();
}
void ExtractTask::extract(SentenceAlignment &sentence)
{
int countE = sentence.target.size();
int countF = sentence.source.size();
HPhraseVector inboundPhrases;
HSentenceVertices inTopLeft;
HSentenceVertices inTopRight;
HSentenceVertices inBottomLeft;
HSentenceVertices inBottomRight;
HSentenceVertices outTopLeft;
HSentenceVertices outTopRight;
HSentenceVertices outBottomLeft;
HSentenceVertices outBottomRight;
HSentenceVertices::const_iterator it;
bool relaxLimit = m_options.isHierModel();
bool buildExtraStructure = m_options.isPhraseModel() || m_options.isHierModel();
// check alignments for target phrase startE...endE
// loop over extracted phrases which are compatible with the word-alignments
for(int startE=0; startE<countE; startE++) {
for(int endE=startE;
(endE<countE && (relaxLimit || endE<startE+m_options.maxPhraseLength));
endE++) {
int minF = 9999;
int maxF = -1;
vector< int > usedF = sentence.alignedCountS;
for(int ei=startE; ei<=endE; ei++) {
for(size_t i=0; i<sentence.alignedToT[ei].size(); i++) {
int fi = sentence.alignedToT[ei][i];
if (fi<minF) {
minF = fi;
}
if (fi>maxF) {
maxF = fi;
}
usedF[ fi ]--;
}
}
if (maxF >= 0 && // aligned to any source words at all
(relaxLimit || maxF-minF < m_options.maxPhraseLength)) { // source phrase within limits
// check if source words are aligned to out of bound target words
bool out_of_bounds = false;
for(int fi=minF; fi<=maxF && !out_of_bounds; fi++)
if (usedF[fi]>0) {
// cout << "ouf of bounds: " << fi << "\n";
out_of_bounds = true;
}
// cout << "doing if for ( " << minF << "-" << maxF << ", " << startE << "," << endE << ")\n";
if (!out_of_bounds) {
// start point of source phrase may retreat over unaligned
for(int startF=minF;
(startF>=0 &&
(relaxLimit || startF>maxF-m_options.maxPhraseLength) && // within length limit
(startF==minF || sentence.alignedCountS[startF]==0)); // unaligned
startF--)
// end point of source phrase may advance over unaligned
for(int endF=maxF;
(endF<countF &&
(relaxLimit || endF<startF+m_options.maxPhraseLength) && // within length limit
(endF==maxF || sentence.alignedCountS[endF]==0)); // unaligned
endF++) { // at this point we have extracted a phrase
if(buildExtraStructure) { // phrase || hier
if(endE-startE < m_options.maxPhraseLength && endF-startF < m_options.maxPhraseLength) { // within limit
inboundPhrases.push_back(HPhrase(HPhraseVertex(startF,startE),
HPhraseVertex(endF,endE)));
insertPhraseVertices(inTopLeft, inTopRight, inBottomLeft, inBottomRight,
startF, startE, endF, endE);
} else
insertPhraseVertices(outTopLeft, outTopRight, outBottomLeft, outBottomRight,
startF, startE, endF, endE);
} else {
string orientationInfo = "";
if(m_options.isWordModel()) {
REO_POS wordPrevOrient, wordNextOrient;
bool connectedLeftTopP = isAligned( sentence, startF-1, startE-1 );
bool connectedRightTopP = isAligned( sentence, endF+1, startE-1 );
bool connectedLeftTopN = isAligned( sentence, endF+1, endE+1 );
bool connectedRightTopN = isAligned( sentence, startF-1, endE+1 );
wordPrevOrient = getOrientWordModel(sentence, m_options.isWordType(), connectedLeftTopP, connectedRightTopP, startF, endF, startE, endE, countF, 0, 1, &ge, <);
wordNextOrient = getOrientWordModel(sentence, m_options.isWordType(), connectedLeftTopN, connectedRightTopN, endF, startF, endE, startE, 0, countF, -1, <, &ge);
orientationInfo += getOrientString(wordPrevOrient, m_options.isWordType()) + " " + getOrientString(wordNextOrient, m_options.isWordType());
if(m_options.isAllModelsOutputFlag())
" | | ";
}
addPhrase(sentence, startE, endE, startF, endF, orientationInfo);
}
}
}
}
}
}
if(buildExtraStructure) { // phrase || hier
string orientationInfo = "";
REO_POS wordPrevOrient, wordNextOrient, phrasePrevOrient, phraseNextOrient, hierPrevOrient, hierNextOrient;
for(size_t i = 0; i < inboundPhrases.size(); i++) {
int startF = inboundPhrases[i].first.first;
int startE = inboundPhrases[i].first.second;
int endF = inboundPhrases[i].second.first;
int endE = inboundPhrases[i].second.second;
bool connectedLeftTopP = isAligned( sentence, startF-1, startE-1 );
bool connectedRightTopP = isAligned( sentence, endF+1, startE-1 );
bool connectedLeftTopN = isAligned( sentence, endF+1, endE+1 );
bool connectedRightTopN = isAligned( sentence, startF-1, endE+1 );
if(m_options.isWordModel()) {
wordPrevOrient = getOrientWordModel(sentence, m_options.isWordType(),
connectedLeftTopP, connectedRightTopP,
startF, endF, startE, endE, countF, 0, 1,
&ge, <);
wordNextOrient = getOrientWordModel(sentence, m_options.isWordType(),
connectedLeftTopN, connectedRightTopN,
endF, startF, endE, startE, 0, countF, -1,
<, &ge);
}
if (m_options.isPhraseModel()) {
phrasePrevOrient = getOrientPhraseModel(sentence, m_options.isPhraseType(),
connectedLeftTopP, connectedRightTopP,
startF, endF, startE, endE, countF-1, 0, 1, &ge, <, inBottomRight, inBottomLeft);
phraseNextOrient = getOrientPhraseModel(sentence, m_options.isPhraseType(),
connectedLeftTopN, connectedRightTopN,
endF, startF, endE, startE, 0, countF-1, -1, <, &ge, inBottomLeft, inBottomRight);
} else {
phrasePrevOrient = phraseNextOrient = UNKNOWN;
}
if(m_options.isHierModel()) {
hierPrevOrient = getOrientHierModel(sentence, m_options.isHierType(),
connectedLeftTopP, connectedRightTopP,
startF, endF, startE, endE, countF-1, 0, 1, &ge, <, inBottomRight, inBottomLeft, outBottomRight, outBottomLeft, phrasePrevOrient);
hierNextOrient = getOrientHierModel(sentence, m_options.isHierType(),
connectedLeftTopN, connectedRightTopN,
endF, startF, endE, startE, 0, countF-1, -1, <, &ge, inBottomLeft, inBottomRight, outBottomLeft, outBottomRight, phraseNextOrient);
}
orientationInfo = ((m_options.isWordModel())? getOrientString(wordPrevOrient, m_options.isWordType()) + " " + getOrientString(wordNextOrient, m_options.isWordType()) : "") + " | " +
((m_options.isPhraseModel())? getOrientString(phrasePrevOrient, m_options.isPhraseType()) + " " + getOrientString(phraseNextOrient, m_options.isPhraseType()) : "") + " | " +
((m_options.isHierModel())? getOrientString(hierPrevOrient, m_options.isHierType()) + " " + getOrientString(hierNextOrient, m_options.isHierType()) : "");
addPhrase(sentence, startE, endE, startF, endF, orientationInfo);
}
}
}
REO_POS getOrientWordModel(SentenceAlignment & sentence, REO_MODEL_TYPE modelType,
bool connectedLeftTop, bool connectedRightTop,
int startF, int endF, int startE, int endE, int countF, int zero, int unit,
bool (*ge)(int, int), bool (*lt)(int, int) )
{
if( connectedLeftTop && !connectedRightTop)
return LEFT;
if(modelType == REO_MONO)
return UNKNOWN;
if (!connectedLeftTop && connectedRightTop)
return RIGHT;
if(modelType == REO_MSD)
return UNKNOWN;
for(int indexF=startF-2*unit; (*ge)(indexF, zero) && !connectedLeftTop; indexF=indexF-unit)
connectedLeftTop = isAligned(sentence, indexF, startE-unit);
for(int indexF=endF+2*unit; (*lt)(indexF,countF) && !connectedRightTop; indexF=indexF+unit)
connectedRightTop = isAligned(sentence, indexF, startE-unit);
if(connectedLeftTop && !connectedRightTop)
return DRIGHT;
else if(!connectedLeftTop && connectedRightTop)
return DLEFT;
return UNKNOWN;
}
// to be called with countF-1 instead of countF
REO_POS getOrientPhraseModel (SentenceAlignment & sentence, REO_MODEL_TYPE modelType,
bool connectedLeftTop, bool connectedRightTop,
int startF, int endF, int startE, int endE, int countF, int zero, int unit,
bool (*ge)(int, int), bool (*lt)(int, int),
const HSentenceVertices & inBottomRight, const HSentenceVertices & inBottomLeft)
{
HSentenceVertices::const_iterator it;
if((connectedLeftTop && !connectedRightTop) ||
//(startE == 0 && startF == 0) ||
//(startE == sentence.target.size()-1 && startF == sentence.source.size()-1) ||
((it = inBottomRight.find(startE - unit)) != inBottomRight.end() &&
it->second.find(startF-unit) != it->second.end()))
return LEFT;
if(modelType == REO_MONO)
return UNKNOWN;
if((!connectedLeftTop && connectedRightTop) ||
((it = inBottomLeft.find(startE - unit)) != inBottomLeft.end() && it->second.find(endF + unit) != it->second.end()))
return RIGHT;
if(modelType == REO_MSD)
return UNKNOWN;
connectedLeftTop = false;
for(int indexF=startF-2*unit; (*ge)(indexF, zero) && !connectedLeftTop; indexF=indexF-unit)
if(connectedLeftTop = (it = inBottomRight.find(startE - unit)) != inBottomRight.end() &&
it->second.find(indexF) != it->second.end())
return DRIGHT;
connectedRightTop = false;
for(int indexF=endF+2*unit; (*lt)(indexF, countF) && !connectedRightTop; indexF=indexF+unit)
if(connectedRightTop = (it = inBottomLeft.find(startE - unit)) != inBottomRight.end() &&
it->second.find(indexF) != it->second.end())
return DLEFT;
return UNKNOWN;
}
// to be called with countF-1 instead of countF
REO_POS getOrientHierModel (SentenceAlignment & sentence, REO_MODEL_TYPE modelType,
bool connectedLeftTop, bool connectedRightTop,
int startF, int endF, int startE, int endE, int countF, int zero, int unit,
bool (*ge)(int, int), bool (*lt)(int, int),
const HSentenceVertices & inBottomRight, const HSentenceVertices & inBottomLeft,
const HSentenceVertices & outBottomRight, const HSentenceVertices & outBottomLeft,
REO_POS phraseOrient)
{
HSentenceVertices::const_iterator it;
if(phraseOrient == LEFT ||
(connectedLeftTop && !connectedRightTop) ||
// (startE == 0 && startF == 0) ||
//(startE == sentence.target.size()-1 && startF == sentence.source.size()-1) ||
((it = inBottomRight.find(startE - unit)) != inBottomRight.end() &&
it->second.find(startF-unit) != it->second.end()) ||
((it = outBottomRight.find(startE - unit)) != outBottomRight.end() &&
it->second.find(startF-unit) != it->second.end()))
return LEFT;
if(modelType == REO_MONO)
return UNKNOWN;
if(phraseOrient == RIGHT ||
(!connectedLeftTop && connectedRightTop) ||
((it = inBottomLeft.find(startE - unit)) != inBottomLeft.end() &&
it->second.find(endF + unit) != it->second.end()) ||
((it = outBottomLeft.find(startE - unit)) != outBottomLeft.end() &&
it->second.find(endF + unit) != it->second.end()))
return RIGHT;
if(modelType == REO_MSD)
return UNKNOWN;
if(phraseOrient != UNKNOWN)
return phraseOrient;
connectedLeftTop = false;
for(int indexF=startF-2*unit; (*ge)(indexF, zero) && !connectedLeftTop; indexF=indexF-unit) {
if((connectedLeftTop = (it = inBottomRight.find(startE - unit)) != inBottomRight.end() &&
it->second.find(indexF) != it->second.end()) ||
(connectedLeftTop = (it = outBottomRight.find(startE - unit)) != outBottomRight.end() &&
it->second.find(indexF) != it->second.end()))
return DRIGHT;
}
connectedRightTop = false;
for(int indexF=endF+2*unit; (*lt)(indexF, countF) && !connectedRightTop; indexF=indexF+unit) {
if((connectedRightTop = (it = inBottomLeft.find(startE - unit)) != inBottomRight.end() &&
it->second.find(indexF) != it->second.end()) ||
(connectedRightTop = (it = outBottomLeft.find(startE - unit)) != outBottomRight.end() &&
it->second.find(indexF) != it->second.end()))
return DLEFT;
}
return UNKNOWN;
}
bool isAligned ( SentenceAlignment &sentence, int fi, int ei )
{
if (ei == -1 && fi == -1)
return true;
if (ei <= -1 || fi <= -1)
return false;
if ((size_t)ei == sentence.target.size() && (size_t)fi == sentence.source.size())
return true;
if ((size_t)ei >= sentence.target.size() || (size_t)fi >= sentence.source.size())
return false;
for(size_t i=0; i<sentence.alignedToT[ei].size(); i++)
if (sentence.alignedToT[ei][i] == fi)
return true;
return false;
}
bool ge(int first, int second)
{
return first >= second;
}
bool le(int first, int second)
{
return first <= second;
}
bool lt(int first, int second)
{
return first < second;
}
void insertVertex( HSentenceVertices & corners, int x, int y )
{
set<int> tmp;
tmp.insert(x);
pair< HSentenceVertices::iterator, bool > ret = corners.insert( pair<int, set<int> > (y, tmp) );
if(ret.second == false) {
ret.first->second.insert(x);
}
}
void insertPhraseVertices(
HSentenceVertices & topLeft,
HSentenceVertices & topRight,
HSentenceVertices & bottomLeft,
HSentenceVertices & bottomRight,
int startF, int startE, int endF, int endE)
{
insertVertex(topLeft, startF, startE);
insertVertex(topRight, endF, startE);
insertVertex(bottomLeft, startF, endE);
insertVertex(bottomRight, endF, endE);
}
string getOrientString(REO_POS orient, REO_MODEL_TYPE modelType)
{
switch(orient) {
case LEFT:
return "mono";
break;
case RIGHT:
return "swap";
break;
case DRIGHT:
return "dright";
break;
case DLEFT:
return "dleft";
break;
case UNKNOWN:
switch(modelType) {
case REO_MONO:
return "nomono";
break;
case REO_MSD:
return "other";
break;
case REO_MSLR:
return "dright";
break;
}
break;
}
return "";
}
void ExtractTask::addPhrase( SentenceAlignment &sentence, int startE, int endE, int startF, int endF , string &orientationInfo)
{
// source
// // cout << "adding ( " << startF << "-" << endF << ", " << startE << "-" << endE << ")\n";
ostringstream outextractstr;
ostringstream outextractstrInv;
ostringstream outextractstrOrientation;
if (m_options.isOnlyOutputSpanInfo()) {
cout << startF << " " << endF << " " << startE << " " << endE << endl;
return;
}
for(int fi=startF; fi<=endF; fi++) {
if (m_options.isTranslationFlag()) outextractstr << sentence.source[fi] << " ";
if (m_options.isOrientationFlag()) outextractstrOrientation << sentence.source[fi] << " ";
}
if (m_options.isTranslationFlag()) outextractstr << "||| ";
if (m_options.isOrientationFlag()) outextractstrOrientation << "||| ";
// target
for(int ei=startE; ei<=endE; ei++) {
if (m_options.isTranslationFlag()) outextractstr << sentence.target[ei] << " ";
if (m_options.isTranslationFlag()) outextractstrInv << sentence.target[ei] << " ";
if (m_options.isOrientationFlag()) outextractstrOrientation << sentence.target[ei] << " ";
}
if (m_options.isTranslationFlag()) outextractstr << "|||";
if (m_options.isTranslationFlag()) outextractstrInv << "||| ";
if (m_options.isOrientationFlag()) outextractstrOrientation << "||| ";
// source (for inverse)
if (m_options.isTranslationFlag()) {
for(int fi=startF; fi<=endF; fi++)
outextractstrInv << sentence.source[fi] << " ";
outextractstrInv << "|||";
}
// alignment
if (m_options.isTranslationFlag()) {
for(int ei=startE; ei<=endE; ei++) {
for(unsigned int i=0; i<sentence.alignedToT[ei].size(); i++) {
int fi = sentence.alignedToT[ei][i];
outextractstr << " " << fi-startF << "-" << ei-startE;
outextractstrInv << " " << ei-startE << "-" << fi-startF;
}
}
}
if (m_options.isOrientationFlag())
outextractstrOrientation << orientationInfo;
if (m_options.getInstanceWeightsFile().length()) {
if (m_options.isTranslationFlag()) {
outextractstr << " ||| " << sentence.weightString;
outextractstrInv << " ||| " << sentence.weightString;
}
if (m_options.isOrientationFlag()) {
outextractstrOrientation << " ||| " << sentence.weightString;
}
}
if (m_options.isIncludeSentenceIdFlag()) {
outextractstr << " ||| " << sentence.sentenceID;
}
if (m_options.isTranslationFlag()) outextractstr << "\n";
if (m_options.isTranslationFlag()) outextractstrInv << "\n";
if (m_options.isOrientationFlag()) outextractstrOrientation << "\n";
m_extractedPhrases.push_back(outextractstr.str());
m_extractedPhrasesInv.push_back(outextractstrInv.str());
m_extractedPhrasesOri.push_back(outextractstrOrientation.str());
}
void ExtractTask::writePhrasesToFile(){
ostringstream outextractFile;
ostringstream outextractFileInv;
ostringstream outextractFileOrientation;
for(vector<string>::const_iterator phrase=m_extractedPhrases.begin();phrase!=m_extractedPhrases.end();phrase++){
outextractFile<<phrase->data();
}
for(vector<string>::const_iterator phrase=m_extractedPhrasesInv.begin();phrase!=m_extractedPhrasesInv.end();phrase++){
outextractFileInv<<phrase->data();
}
for(vector<string>::const_iterator phrase=m_extractedPhrasesOri.begin();phrase!=m_extractedPhrasesOri.end();phrase++){
outextractFileOrientation<<phrase->data();
}
m_extractFile << outextractFile.str();
m_extractFileInv << outextractFileInv.str();
m_extractFileOrientation << outextractFileOrientation.str();
}
// if proper conditioning, we need the number of times a source phrase occured
void ExtractTask::extractBase( SentenceAlignment &sentence )
{
ostringstream outextractFile;
ostringstream outextractFileInv;
int countF = sentence.source.size();
for(int startF=0; startF<countF; startF++) {
for(int endF=startF;
(endF<countF && endF<startF+m_options.maxPhraseLength);
endF++) {
for(int fi=startF; fi<=endF; fi++) {
outextractFile << sentence.source[fi] << " ";
}
outextractFile << "|||" << endl;
}
}
int countE = sentence.target.size();
for(int startE=0; startE<countE; startE++) {
for(int endE=startE;
(endE<countE && endE<startE+m_options.maxPhraseLength);
endE++) {
for(int ei=startE; ei<=endE; ei++) {
outextractFileInv << sentence.target[ei] << " ";
}
outextractFileInv << "|||" << endl;
}
}
m_extractFile << outextractFile.str();
m_extractFileInv << outextractFileInv.str();
}
}
|