Hot-keys on this page

r m x p   toggle line displays

j k   next/prev highlighted chunk

0   (zero) top of page

1   (one) first highlighted chunk

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

797

798

799

800

801

802

803

804

805

806

807

808

809

810

811

812

813

814

815

816

817

818

819

820

821

822

823

824

825

826

827

828

829

830

831

832

833

834

835

836

837

838

839

840

841

842

843

844

845

846

847

848

849

850

851

852

853

854

855

856

857

858

859

860

861

862

863

864

865

866

867

868

869

870

871

872

873

874

875

876

877

878

879

880

881

882

883

884

885

886

887

888

889

890

891

892

893

894

895

896

897

898

899

900

901

902

903

904

905

906

907

908

909

910

911

912

913

914

915

916

917

918

919

920

921

922

923

924

925

926

927

928

929

930

931

932

933

934

935

936

937

938

939

940

941

942

943

944

945

946

947

948

949

950

951

952

953

954

955

956

957

958

959

960

961

962

963

964

965

966

967

968

969

970

971

972

973

974

975

976

977

978

979

980

981

982

983

984

985

986

987

988

989

990

991

992

993

994

995

996

997

998

999

1000

1001

1002

1003

1004

1005

1006

1007

1008

1009

1010

1011

1012

1013

1014

1015

1016

1017

1018

1019

1020

1021

1022

1023

1024

1025

1026

1027

1028

1029

1030

1031

1032

1033

1034

1035

1036

1037

1038

1039

1040

1041

1042

1043

1044

1045

1046

1047

1048

1049

1050

1051

1052

1053

1054

1055

1056

1057

1058

1059

1060

1061

1062

1063

1064

1065

1066

1067

1068

1069

1070

1071

1072

1073

1074

1075

1076

1077

1078

1079

1080

1081

1082

1083

1084

1085

1086

1087

1088

1089

1090

1091

1092

1093

1094

1095

1096

1097

1098

1099

1100

1101

1102

1103

1104

1105

1106

1107

1108

1109

1110

1111

1112

1113

1114

1115

1116

1117

1118

1119

1120

1121

1122

1123

1124

1125

1126

1127

1128

1129

1130

1131

1132

1133

1134

1135

1136

1137

1138

1139

1140

1141

1142

1143

1144

1145

1146

1147

1148

1149

1150

1151

1152

1153

1154

1155

1156

1157

1158

1159

1160

1161

1162

1163

1164

1165

1166

1167

1168

1169

1170

1171

1172

1173

1174

1175

1176

1177

1178

1179

1180

1181

1182

1183

1184

1185

1186

1187

1188

1189

1190

1191

1192

1193

1194

1195

1196

1197

1198

1199

1200

1201

1202

1203

1204

1205

1206

1207

1208

1209

1210

1211

1212

1213

1214

1215

1216

1217

1218

1219

1220

1221

1222

1223

1224

1225

1226

1227

1228

1229

1230

1231

1232

1233

1234

1235

1236

1237

1238

1239

1240

1241

1242

1243

1244

1245

1246

1247

1248

1249

1250

1251

1252

1253

1254

1255

1256

1257

1258

1259

1260

1261

1262

1263

1264

1265

1266

1267

1268

1269

1270

1271

1272

1273

1274

1275

1276

1277

1278

1279

1280

1281

1282

1283

1284

1285

1286

1287

1288

1289

1290

1291

1292

1293

1294

1295

1296

1297

1298

1299

1300

1301

1302

1303

1304

1305

1306

1307

1308

1309

1310

1311

1312

1313

1314

1315

1316

1317

1318

1319

1320

1321

1322

1323

1324

1325

1326

1327

1328

1329

1330

1331

1332

1333

1334

1335

1336

1337

1338

1339

1340

1341

1342

1343

1344

1345

1346

1347

1348

1349

1350

1351

1352

1353

1354

1355

1356

1357

1358

1359

1360

1361

1362

1363

1364

1365

1366

1367

1368

1369

1370

1371

1372

1373

1374

1375

1376

1377

1378

1379

1380

1381

1382

1383

1384

1385

1386

1387

1388

1389

1390

1391

1392

1393

1394

1395

1396

1397

1398

1399

1400

1401

1402

1403

1404

1405

1406

1407

1408

1409

1410

1411

1412

1413

1414

1415

1416

1417

1418

1419

1420

1421

1422

1423

1424

1425

1426

1427

1428

1429

1430

1431

1432

1433

1434

1435

1436

1437

1438

1439

1440

1441

1442

1443

1444

1445

1446

1447

1448

1449

1450

1451

1452

1453

1454

1455

1456

1457

1458

1459

1460

1461

1462

1463

1464

1465

1466

1467

1468

1469

1470

1471

1472

1473

1474

1475

1476

1477

1478

1479

1480

1481

1482

1483

1484

1485

1486

1487

1488

1489

1490

1491

1492

1493

1494

1495

1496

1497

1498

1499

1500

1501

1502

1503

1504

1505

1506

1507

1508

1509

1510

1511

1512

1513

1514

1515

1516

1517

1518

1519

1520

1521

1522

1523

1524

1525

1526

1527

1528

1529

1530

1531

1532

1533

1534

1535

1536

1537

1538

1539

1540

1541

1542

1543

1544

1545

1546

1547

1548

1549

1550

1551

1552

1553

1554

1555

1556

1557

1558

1559

1560

1561

1562

1563

1564

1565

1566

1567

1568

1569

1570

1571

1572

1573

1574

1575

1576

1577

1578

1579

1580

1581

1582

1583

1584

1585

1586

1587

1588

1589

1590

1591

1592

1593

1594

1595

1596

1597

1598

1599

1600

1601

1602

1603

1604

1605

1606

1607

1608

1609

1610

1611

1612

1613

1614

1615

1616

1617

1618

1619

1620

1621

1622

1623

1624

1625

1626

1627

1628

1629

1630

1631

1632

1633

1634

1635

1636

1637

1638

1639

1640

1641

1642

1643

1644

1645

1646

1647

1648

1649

1650

1651

1652

1653

1654

1655

1656

1657

1658

1659

1660

1661

1662

1663

1664

1665

1666

1667

1668

1669

1670

1671

1672

1673

1674

1675

1676

1677

1678

1679

1680

1681

1682

1683

1684

1685

1686

1687

1688

1689

1690

1691

1692

1693

1694

1695

1696

1697

1698

1699

1700

1701

1702

1703

1704

1705

1706

1707

1708

1709

1710

1711

1712

1713

1714

1715

1716

1717

1718

1719

1720

1721

1722

1723

1724

1725

1726

1727

1728

1729

# -*- coding: utf-8 -*- 

# Natural Language Toolkit: A Chart Parser 

# 

# Copyright (C) 2001-2012 NLTK Project 

# Author: Edward Loper <edloper@gradient.cis.upenn.edu> 

#         Steven Bird <sb@csse.unimelb.edu.au> 

#         Jean Mark Gawron <gawron@mail.sdsu.edu> 

#         Peter Ljunglöf <peter.ljunglof@heatherleaf.se> 

# URL: <http://www.nltk.org/> 

# For license information, see LICENSE.TXT 

 

""" 

Data classes and parser implementations for "chart parsers", which 

use dynamic programming to efficiently parse a text.  A chart 

parser derives parse trees for a text by iteratively adding "edges" 

to a "chart."  Each edge represents a hypothesis about the tree 

structure for a subsequence of the text.  The chart is a 

"blackboard" for composing and combining these hypotheses. 

 

When a chart parser begins parsing a text, it creates a new (empty) 

chart, spanning the text.  It then incrementally adds new edges to the 

chart.  A set of "chart rules" specifies the conditions under which 

new edges should be added to the chart.  Once the chart reaches a 

stage where none of the chart rules adds any new edges, parsing is 

complete. 

 

Charts are encoded with the ``Chart`` class, and edges are encoded with 

the ``TreeEdge`` and ``LeafEdge`` classes.  The chart parser module 

defines three chart parsers: 

 

  - ``ChartParser`` is a simple and flexible chart parser.  Given a 

    set of chart rules, it will apply those rules to the chart until 

    no more edges are added. 

 

  - ``SteppingChartParser`` is a subclass of ``ChartParser`` that can 

    be used to step through the parsing process. 

""" 

from __future__ import print_function, division 

 

import re 

import warnings 

 

from nltk import compat 

from nltk.tree import Tree 

from nltk.grammar import WeightedGrammar, is_nonterminal, is_terminal 

 

from nltk.parse.api import ParserI 

 

 

######################################################################## 

##  Edges 

######################################################################## 

 

class EdgeI(object): 

    """ 

    A hypothesis about the structure of part of a sentence. 

    Each edge records the fact that a structure is (partially) 

    consistent with the sentence.  An edge contains: 

 

    - A span, indicating what part of the sentence is 

      consistent with the hypothesized structure. 

    - A left-hand side, specifying what kind of structure is 

      hypothesized. 

    - A right-hand side, specifying the contents of the 

      hypothesized structure. 

    - A dot position, indicating how much of the hypothesized 

      structure is consistent with the sentence. 

 

    Every edge is either complete or incomplete: 

 

    - An edge is complete if its structure is fully consistent 

      with the sentence. 

    - An edge is incomplete if its structure is partially 

      consistent with the sentence.  For every incomplete edge, the 

      span specifies a possible prefix for the edge's structure. 

 

    There are two kinds of edge: 

 

    - A ``TreeEdge`` records which trees have been found to 

      be (partially) consistent with the text. 

    - A ``LeafEdge`` records the tokens occurring in the text. 

 

    The ``EdgeI`` interface provides a common interface to both types 

    of edge, allowing chart parsers to treat them in a uniform manner. 

    """ 

    def __init__(self): 

        if self.__class__ == EdgeI: 

            raise TypeError('Edge is an abstract interface') 

 

    #//////////////////////////////////////////////////////////// 

    # Span 

    #//////////////////////////////////////////////////////////// 

 

    def span(self): 

        """ 

        Return a tuple ``(s, e)``, where ``tokens[s:e]`` is the 

        portion of the sentence that is consistent with this 

        edge's structure. 

 

        :rtype: tuple(int, int) 

        """ 

        raise NotImplementedError() 

 

    def start(self): 

        """ 

        Return the start index of this edge's span. 

 

        :rtype: int 

        """ 

        raise NotImplementedError() 

 

    def end(self): 

        """ 

        Return the end index of this edge's span. 

 

        :rtype: int 

        """ 

        raise NotImplementedError() 

 

    def length(self): 

        """ 

        Return the length of this edge's span. 

 

        :rtype: int 

        """ 

        raise NotImplementedError() 

 

    #//////////////////////////////////////////////////////////// 

    # Left Hand Side 

    #//////////////////////////////////////////////////////////// 

 

    def lhs(self): 

        """ 

        Return this edge's left-hand side, which specifies what kind 

        of structure is hypothesized by this edge. 

 

        :see: ``TreeEdge`` and ``LeafEdge`` for a description of 

            the left-hand side values for each edge type. 

        """ 

        raise NotImplementedError() 

 

    #//////////////////////////////////////////////////////////// 

    # Right Hand Side 

    #//////////////////////////////////////////////////////////// 

 

    def rhs(self): 

        """ 

        Return this edge's right-hand side, which specifies 

        the content of the structure hypothesized by this edge. 

 

        :see: ``TreeEdge`` and ``LeafEdge`` for a description of 

            the right-hand side values for each edge type. 

        """ 

        raise NotImplementedError() 

 

    def dot(self): 

        """ 

        Return this edge's dot position, which indicates how much of 

        the hypothesized structure is consistent with the 

        sentence.  In particular, ``self.rhs[:dot]`` is consistent 

        with ``tokens[self.start():self.end()]``. 

 

        :rtype: int 

        """ 

        raise NotImplementedError() 

 

    def next(self): 

        """ 

        Return the element of this edge's right-hand side that 

        immediately follows its dot. 

 

        :rtype: Nonterminal or terminal or None 

        """ 

        raise NotImplementedError() 

 

    def __next__(self): 

        return self.next() 

 

    def is_complete(self): 

        """ 

        Return True if this edge's structure is fully consistent 

        with the text. 

 

        :rtype: bool 

        """ 

        raise NotImplementedError() 

 

    def is_incomplete(self): 

        """ 

        Return True if this edge's structure is partially consistent 

        with the text. 

 

        :rtype: bool 

        """ 

        raise NotImplementedError() 

 

    #//////////////////////////////////////////////////////////// 

    # Comparisons 

    #//////////////////////////////////////////////////////////// 

 

    def __cmp__(self, other): 

        raise NotImplementedError() 

 

    def __hash__(self, other): 

        raise NotImplementedError() 

 

class TreeEdge(EdgeI): 

    """ 

    An edge that records the fact that a tree is (partially) 

    consistent with the sentence.  A tree edge consists of: 

 

    - A span, indicating what part of the sentence is 

      consistent with the hypothesized tree. 

    - A left-hand side, specifying the hypothesized tree's node 

      value. 

    - A right-hand side, specifying the hypothesized tree's 

      children.  Each element of the right-hand side is either a 

      terminal, specifying a token with that terminal as its leaf 

      value; or a nonterminal, specifying a subtree with that 

      nonterminal's symbol as its node value. 

    - A dot position, indicating which children are consistent 

      with part of the sentence.  In particular, if ``dot`` is the 

      dot position, ``rhs`` is the right-hand size, ``(start,end)`` 

      is the span, and ``sentence`` is the list of tokens in the 

      sentence, then ``tokens[start:end]`` can be spanned by the 

      children specified by ``rhs[:dot]``. 

 

    For more information about edges, see the ``EdgeI`` interface. 

    """ 

    def __init__(self, span, lhs, rhs, dot=0): 

        """ 

        Construct a new ``TreeEdge``. 

 

        :type span: tuple(int, int) 

        :param span: A tuple ``(s, e)``, where ``tokens[s:e]`` is the 

            portion of the sentence that is consistent with the new 

            edge's structure. 

        :type lhs: Nonterminal 

        :param lhs: The new edge's left-hand side, specifying the 

            hypothesized tree's node value. 

        :type rhs: list(Nonterminal and str) 

        :param rhs: The new edge's right-hand side, specifying the 

            hypothesized tree's children. 

        :type dot: int 

        :param dot: The position of the new edge's dot.  This position 

            specifies what prefix of the production's right hand side 

            is consistent with the text.  In particular, if 

            ``sentence`` is the list of tokens in the sentence, then 

            ``okens[span[0]:span[1]]`` can be spanned by the 

            children specified by ``rhs[:dot]``. 

        """ 

        self._lhs = lhs 

        self._rhs = tuple(rhs) 

        self._span = span 

        self._dot = dot 

 

    @staticmethod 

    def from_production(production, index): 

        """ 

        Return a new ``TreeEdge`` formed from the given production. 

        The new edge's left-hand side and right-hand side will 

        be taken from ``production``; its span will be 

        ``(index,index)``; and its dot position will be ``0``. 

 

        :rtype: TreeEdge 

        """ 

        return TreeEdge(span=(index, index), lhs=production.lhs(), 

                        rhs=production.rhs(), dot=0) 

 

    def move_dot_forward(self, new_end): 

        """ 

        Return a new ``TreeEdge`` formed from this edge. 

        The new edge's dot position is increased by ``1``, 

        and its end index will be replaced by ``new_end``. 

 

        :param new_end: The new end index. 

        :type new_end: int 

        :rtype: TreeEdge 

        """ 

        return TreeEdge(span=(self._span[0], new_end), 

                        lhs=self._lhs, rhs=self._rhs, 

                        dot=self._dot+1) 

 

    # Accessors 

    def lhs(self): return self._lhs 

    def span(self): return self._span 

    def start(self): return self._span[0] 

    def end(self): return self._span[1] 

    def length(self): return self._span[1] - self._span[0] 

    def rhs(self): return self._rhs 

    def dot(self): return self._dot 

    def is_complete(self): return self._dot == len(self._rhs) 

    def is_incomplete(self): return self._dot != len(self._rhs) 

    def next(self): 

        if self._dot >= len(self._rhs): return None 

        else: return self._rhs[self._dot] 

 

    # Comparisons & hashing 

    def __cmp__(self, other): 

        if self.__class__ != other.__class__: return -1 

        return cmp((self._span, self.lhs(), self.rhs(), self._dot), 

                   (other._span, other.lhs(), other.rhs(), other._dot)) 

    def __hash__(self): 

        return hash((self.lhs(), self.rhs(), self._span, self._dot)) 

 

    # String representation 

    def __str__(self): 

        str = '[%s:%s] ' % (self._span[0], self._span[1]) 

        str += '%-2r ->' % (self._lhs,) 

 

        for i in range(len(self._rhs)): 

            if i == self._dot: str += ' *' 

            str += ' %r' % (self._rhs[i],) 

        if len(self._rhs) == self._dot: str += ' *' 

        return str 

 

    def __repr__(self): 

        return '[Edge: %s]' % self 

 

class LeafEdge(EdgeI): 

    """ 

    An edge that records the fact that a leaf value is consistent with 

    a word in the sentence.  A leaf edge consists of: 

 

    - An index, indicating the position of the word. 

    - A leaf, specifying the word's content. 

 

    A leaf edge's left-hand side is its leaf value, and its right hand 

    side is ``()``.  Its span is ``[index, index+1]``, and its dot 

    position is ``0``. 

    """ 

    def __init__(self, leaf, index): 

        """ 

        Construct a new ``LeafEdge``. 

 

        :param leaf: The new edge's leaf value, specifying the word 

            that is recorded by this edge. 

        :param index: The new edge's index, specifying the position of 

            the word that is recorded by this edge. 

        """ 

        self._leaf = leaf 

        self._index = index 

 

    # Accessors 

    def lhs(self): return self._leaf 

    def span(self): return (self._index, self._index+1) 

    def start(self): return self._index 

    def end(self): return self._index+1 

    def length(self): return 1 

    def rhs(self): return () 

    def dot(self): return 0 

    def is_complete(self): return True 

    def is_incomplete(self): return False 

    def next(self): return None 

 

    # Comparisons & hashing 

    def __cmp__(self, other): 

        if not isinstance(other, LeafEdge): return -1 

        return cmp((self._index, self._leaf), (other._index, other._leaf)) 

    def __hash__(self): 

        return hash((self._index, self._leaf)) 

 

    # String representations 

    def __str__(self): 

        return '[%s:%s] %r' % (self._index, self._index+1, self._leaf) 

    def __repr__(self): 

        return '[Edge: %s]' % (self) 

 

######################################################################## 

##  Chart 

######################################################################## 

 

class Chart(object): 

    """ 

    A blackboard for hypotheses about the syntactic constituents of a 

    sentence.  A chart contains a set of edges, and each edge encodes 

    a single hypothesis about the structure of some portion of the 

    sentence. 

 

    The ``select`` method can be used to select a specific collection 

    of edges.  For example ``chart.select(is_complete=True, start=0)`` 

    yields all complete edges whose start indices are 0.  To ensure 

    the efficiency of these selection operations, ``Chart`` dynamically 

    creates and maintains an index for each set of attributes that 

    have been selected on. 

 

    In order to reconstruct the trees that are represented by an edge, 

    the chart associates each edge with a set of child pointer lists. 

    A child pointer list is a list of the edges that license an 

    edge's right-hand side. 

 

    :ivar _tokens: The sentence that the chart covers. 

    :ivar _num_leaves: The number of tokens. 

    :ivar _edges: A list of the edges in the chart 

    :ivar _edge_to_cpls: A dictionary mapping each edge to a set 

        of child pointer lists that are associated with that edge. 

    :ivar _indexes: A dictionary mapping tuples of edge attributes 

        to indices, where each index maps the corresponding edge 

        attribute values to lists of edges. 

    """ 

    def __init__(self, tokens): 

        """ 

        Construct a new chart. The chart is initialized with the 

        leaf edges corresponding to the terminal leaves. 

 

        :type tokens: list 

        :param tokens: The sentence that this chart will be used to parse. 

        """ 

        # Record the sentence token and the sentence length. 

        self._tokens = tuple(tokens) 

        self._num_leaves = len(self._tokens) 

 

        # Initialise the chart. 

        self.initialize() 

 

    def initialize(self): 

        """ 

        Clear the chart. 

        """ 

        # A list of edges contained in this chart. 

        self._edges = [] 

 

        # The set of child pointer lists associated with each edge. 

        self._edge_to_cpls = {} 

 

        # Indexes mapping attribute values to lists of edges 

        # (used by select()). 

        self._indexes = {} 

 

    #//////////////////////////////////////////////////////////// 

    # Sentence Access 

    #//////////////////////////////////////////////////////////// 

 

    def num_leaves(self): 

        """ 

        Return the number of words in this chart's sentence. 

 

        :rtype: int 

        """ 

        return self._num_leaves 

 

    def leaf(self, index): 

        """ 

        Return the leaf value of the word at the given index. 

 

        :rtype: str 

        """ 

        return self._tokens[index] 

 

    def leaves(self): 

        """ 

        Return a list of the leaf values of each word in the 

        chart's sentence. 

 

        :rtype: list(str) 

        """ 

        return self._tokens 

 

    #//////////////////////////////////////////////////////////// 

    # Edge access 

    #//////////////////////////////////////////////////////////// 

 

    def edges(self): 

        """ 

        Return a list of all edges in this chart.  New edges 

        that are added to the chart after the call to edges() 

        will *not* be contained in this list. 

 

        :rtype: list(EdgeI) 

        :see: ``iteredges``, ``select`` 

        """ 

        return self._edges[:] 

 

    def iteredges(self): 

        """ 

        Return an iterator over the edges in this chart.  It is 

        not guaranteed that new edges which are added to the 

        chart before the iterator is exhausted will also be generated. 

 

        :rtype: iter(EdgeI) 

        :see: ``edges``, ``select`` 

        """ 

        return iter(self._edges) 

 

    # Iterating over the chart yields its edges. 

    __iter__ = iteredges 

 

    def num_edges(self): 

        """ 

        Return the number of edges contained in this chart. 

 

        :rtype: int 

        """ 

        return len(self._edge_to_cpls) 

 

    def select(self, **restrictions): 

        """ 

        Return an iterator over the edges in this chart.  Any 

        new edges that are added to the chart before the iterator 

        is exahusted will also be generated.  ``restrictions`` 

        can be used to restrict the set of edges that will be 

        generated. 

 

        :param span: Only generate edges ``e`` where ``e.span()==span`` 

        :param start: Only generate edges ``e`` where ``e.start()==start`` 

        :param end: Only generate edges ``e`` where ``e.end()==end`` 

        :param length: Only generate edges ``e`` where ``e.length()==length`` 

        :param lhs: Only generate edges ``e`` where ``e.lhs()==lhs`` 

        :param rhs: Only generate edges ``e`` where ``e.rhs()==rhs`` 

        :param next: Only generate edges ``e`` where ``e.next()==next`` 

        :param dot: Only generate edges ``e`` where ``e.dot()==dot`` 

        :param is_complete: Only generate edges ``e`` where 

            ``e.is_complete()==is_complete`` 

        :param is_incomplete: Only generate edges ``e`` where 

            ``e.is_incomplete()==is_incomplete`` 

        :rtype: iter(EdgeI) 

        """ 

        # If there are no restrictions, then return all edges. 

        if restrictions=={}: return iter(self._edges) 

 

        # Find the index corresponding to the given restrictions. 

        restr_keys = sorted(restrictions.keys()) 

        restr_keys = tuple(restr_keys) 

 

        # If it doesn't exist, then create it. 

        if restr_keys not in self._indexes: 

            self._add_index(restr_keys) 

 

        vals = tuple(restrictions[key] for key in restr_keys) 

        return iter(self._indexes[restr_keys].get(vals, [])) 

 

    def _add_index(self, restr_keys): 

        """ 

        A helper function for ``select``, which creates a new index for 

        a given set of attributes (aka restriction keys). 

        """ 

        # Make sure it's a valid index. 

        for key in restr_keys: 

            if not hasattr(EdgeI, key): 

                raise ValueError('Bad restriction: %s' % key) 

 

        # Create the index. 

        index = self._indexes[restr_keys] = {} 

 

        # Add all existing edges to the index. 

        for edge in self._edges: 

            vals = tuple(getattr(edge, key)() for key in restr_keys) 

            index.setdefault(vals, []).append(edge) 

 

    def _register_with_indexes(self, edge): 

        """ 

        A helper function for ``insert``, which registers the new 

        edge with all existing indexes. 

        """ 

        for (restr_keys, index) in self._indexes.items(): 

            vals = tuple(getattr(edge, key)() for key in restr_keys) 

            index.setdefault(vals, []).append(edge) 

 

    #//////////////////////////////////////////////////////////// 

    # Edge Insertion 

    #//////////////////////////////////////////////////////////// 

 

    def insert_with_backpointer(self, new_edge, previous_edge, child_edge): 

        """ 

        Add a new edge to the chart, using a pointer to the previous edge. 

        """ 

        cpls = self.child_pointer_lists(previous_edge) 

        new_cpls = [cpl+(child_edge,) for cpl in cpls] 

        return self.insert(new_edge, *new_cpls) 

 

    def insert(self, edge, *child_pointer_lists): 

        """ 

        Add a new edge to the chart, and return True if this operation 

        modified the chart.  In particular, return true iff the chart 

        did not already contain ``edge``, or if it did not already associate 

        ``child_pointer_lists`` with ``edge``. 

 

        :type edge: EdgeI 

        :param edge: The new edge 

        :type child_pointer_lists: sequence of tuple(EdgeI) 

        :param child_pointer_lists: A sequence of lists of the edges that 

            were used to form this edge.  This list is used to reconstruct 

            the trees (or partial trees) that are associated with ``edge``. 

        :rtype: bool 

        """ 

        # Is it a new edge? 

        if edge not in self._edge_to_cpls: 

            # Add it to the list of edges. 

            self._append_edge(edge) 

            # Register with indexes. 

            self._register_with_indexes(edge) 

 

        # Get the set of child pointer lists for this edge. 

        cpls = self._edge_to_cpls.setdefault(edge,{}) 

        chart_was_modified = False 

        for child_pointer_list in child_pointer_lists: 

            child_pointer_list = tuple(child_pointer_list) 

            if child_pointer_list not in cpls: 

                # It's a new CPL; register it, and return true. 

                cpls[child_pointer_list] = True 

                chart_was_modified = True 

        return chart_was_modified 

 

    def _append_edge(self, edge): 

        self._edges.append(edge) 

 

    #//////////////////////////////////////////////////////////// 

    # Tree extraction & child pointer lists 

    #//////////////////////////////////////////////////////////// 

 

    def parses(self, root, tree_class=Tree): 

        """ 

        Return a list of the complete tree structures that span 

        the entire chart, and whose root node is ``root``. 

        """ 

        trees = [] 

        for edge in self.select(start=0, end=self._num_leaves, lhs=root): 

            trees += self.trees(edge, tree_class=tree_class, complete=True) 

        return trees 

 

    def trees(self, edge, tree_class=Tree, complete=False): 

        """ 

        Return a list of the tree structures that are associated 

        with ``edge``. 

 

        If ``edge`` is incomplete, then the unexpanded children will be 

        encoded as childless subtrees, whose node value is the 

        corresponding terminal or nonterminal. 

 

        :rtype: list(Tree) 

        :note: If two trees share a common subtree, then the same 

            Tree may be used to encode that subtree in 

            both trees.  If you need to eliminate this subtree 

            sharing, then create a deep copy of each tree. 

        """ 

        return self._trees(edge, complete, memo={}, tree_class=tree_class) 

 

    def _trees(self, edge, complete, memo, tree_class): 

        """ 

        A helper function for ``trees``. 

 

        :param memo: A dictionary used to record the trees that we've 

            generated for each edge, so that when we see an edge more 

            than once, we can reuse the same trees. 

        """ 

        # If we've seen this edge before, then reuse our old answer. 

        if edge in memo: 

            return memo[edge] 

 

        trees = [] 

 

        # when we're reading trees off the chart, don't use incomplete edges 

        if complete and edge.is_incomplete(): 

            return trees 

 

        # Until we're done computing the trees for edge, set 

        # memo[edge] to be empty.  This has the effect of filtering 

        # out any cyclic trees (i.e., trees that contain themselves as 

        # descendants), because if we reach this edge via a cycle, 

        # then it will appear that the edge doesn't generate any 

        # trees. 

        memo[edge] = [] 

 

        # Leaf edges. 

        if isinstance(edge, LeafEdge): 

            leaf = self._tokens[edge.start()] 

            memo[edge] = leaf 

            return [leaf] 

 

        # Each child pointer list can be used to form trees. 

        for cpl in self.child_pointer_lists(edge): 

            # Get the set of child choices for each child pointer. 

            # child_choices[i] is the set of choices for the tree's 

            # ith child. 

            child_choices = [self._trees(cp, complete, memo, tree_class) 

                             for cp in cpl] 

 

            # For each combination of children, add a tree. 

            for children in self._choose_children(child_choices): 

                lhs = edge.lhs().symbol() 

                trees.append(tree_class(lhs, children)) 

 

        # If the edge is incomplete, then extend it with "partial trees": 

        if edge.is_incomplete(): 

            unexpanded = [tree_class(elt,[]) 

                          for elt in edge.rhs()[edge.dot():]] 

            for tree in trees: 

                tree.extend(unexpanded) 

 

        # Update the memoization dictionary. 

        memo[edge] = trees 

 

        # Return the list of trees. 

        return trees 

 

    def _choose_children(self, child_choices): 

        """ 

        A helper function for ``_trees`` that finds the possible sets 

        of subtrees for a new tree. 

 

        :param child_choices: A list that specifies the options for 

            each child.  In particular, ``child_choices[i]`` is a list of 

            tokens and subtrees that can be used as the ``i``th child. 

        """ 

        children_lists = [[]] 

        for child_choice in child_choices: 

            if hasattr(child_choice, '__iter__') and \ 

                    not isinstance(child_choice, compat.string_types): 

                # Only iterate over the child trees 

                # if child_choice is iterable and NOT a string 

                children_lists = [child_list+[child] 

                                  for child in child_choice 

                                  for child_list in children_lists] 

            else: 

                # If child_choice is a string (or non-iterable) 

                # then it is a leaf 

                children_lists = [child_list+[child_choice] 

                                  for child_list in children_lists] 

        return children_lists 

 

    def child_pointer_lists(self, edge): 

        """ 

        Return the set of child pointer lists for the given edge. 

        Each child pointer list is a list of edges that have 

        been used to form this edge. 

 

        :rtype: list(list(EdgeI)) 

        """ 

        # Make a copy, in case they modify it. 

        return self._edge_to_cpls.get(edge, {}).keys() 

 

    #//////////////////////////////////////////////////////////// 

    # Display 

    #//////////////////////////////////////////////////////////// 

    def pp_edge(self, edge, width=None): 

        """ 

        Return a pretty-printed string representation of a given edge 

        in this chart. 

 

        :rtype: str 

        :param width: The number of characters allotted to each 

            index in the sentence. 

        """ 

        if width is None: width = 50 // (self.num_leaves()+1) 

        (start, end) = (edge.start(), edge.end()) 

 

        str = '|' + ('.'+' '*(width-1))*start 

 

        # Zero-width edges are "#" if complete, ">" if incomplete 

        if start == end: 

            if edge.is_complete(): str += '#' 

            else: str += '>' 

 

        # Spanning complete edges are "[===]"; Other edges are 

        # "[---]" if complete, "[--->" if incomplete 

        elif edge.is_complete() and edge.span() == (0,self._num_leaves): 

            str += '['+('='*width)*(end-start-1) + '='*(width-1)+']' 

        elif edge.is_complete(): 

            str += '['+('-'*width)*(end-start-1) + '-'*(width-1)+']' 

        else: 

            str += '['+('-'*width)*(end-start-1) + '-'*(width-1)+'>' 

 

        str += (' '*(width-1)+'.')*(self._num_leaves-end) 

        return str + '| %s' % edge 

 

    def pp_leaves(self, width=None): 

        """ 

        Return a pretty-printed string representation of this 

        chart's leaves.  This string can be used as a header 

        for calls to ``pp_edge``. 

        """ 

        if width is None: width = 50 // (self.num_leaves()+1) 

 

        if self._tokens is not None and width>1: 

            header = '|.' 

            for tok in self._tokens: 

                header += tok[:width-1].center(width-1)+'.' 

            header += '|' 

        else: 

            header = '' 

 

        return header 

 

    def pp(self, width=None): 

        """ 

        Return a pretty-printed string representation of this chart. 

 

        :param width: The number of characters allotted to each 

            index in the sentence. 

        :rtype: str 

        """ 

        if width is None: width = 50 // (self.num_leaves()+1) 

        # sort edges: primary key=length, secondary key=start index. 

        # (and filter out the token edges) 

        edges = sorted([(e.length(), e.start(), e) for e in self]) 

        edges = [e for (_,_,e) in edges] 

 

        return (self.pp_leaves(width) + '\n' + 

                '\n'.join(self.pp_edge(edge, width) for edge in edges)) 

 

    #//////////////////////////////////////////////////////////// 

    # Display: Dot (AT&T Graphviz) 

    #//////////////////////////////////////////////////////////// 

 

    def dot_digraph(self): 

        # Header 

        s = 'digraph nltk_chart {\n' 

        #s += '  size="5,5";\n' 

        s += '  rankdir=LR;\n' 

        s += '  node [height=0.1,width=0.1];\n' 

        s += '  node [style=filled, color="lightgray"];\n' 

 

        # Set up the nodes 

        for y in range(self.num_edges(), -1, -1): 

            if y == 0: 

                s += '  node [style=filled, color="black"];\n' 

            for x in range(self.num_leaves()+1): 

                if y == 0 or (x <= self._edges[y-1].start() or 

                              x >= self._edges[y-1].end()): 

                    s += '  %04d.%04d [label=""];\n' % (x,y) 

 

        # Add a spacer 

        s += '  x [style=invis]; x->0000.0000 [style=invis];\n' 

 

        # Declare ranks. 

        for x in range(self.num_leaves()+1): 

            s += '  {rank=same;' 

            for y in range(self.num_edges()+1): 

                if y == 0 or (x <= self._edges[y-1].start() or 

                              x >= self._edges[y-1].end()): 

                    s += ' %04d.%04d' % (x,y) 

            s += '}\n' 

 

        # Add the leaves 

        s += '  edge [style=invis, weight=100];\n' 

        s += '  node [shape=plaintext]\n' 

        s += '  0000.0000' 

        for x in range(self.num_leaves()): 

            s += '->%s->%04d.0000' % (self.leaf(x), x+1) 

        s += ';\n\n' 

 

        # Add the edges 

        s += '  edge [style=solid, weight=1];\n' 

        for y, edge in enumerate(self): 

            for x in range(edge.start()): 

                s += ('  %04d.%04d -> %04d.%04d [style="invis"];\n' % 

                      (x, y+1, x+1, y+1)) 

            s += ('  %04d.%04d -> %04d.%04d [label="%s"];\n' % 

                  (edge.start(), y+1, edge.end(), y+1, edge)) 

            for x in range(edge.end(), self.num_leaves()): 

                s += ('  %04d.%04d -> %04d.%04d [style="invis"];\n' % 

                      (x, y+1, x+1, y+1)) 

        s += '}\n' 

        return s 

 

######################################################################## 

##  Chart Rules 

######################################################################## 

 

class ChartRuleI(object): 

    """ 

    A rule that specifies what new edges are licensed by any given set 

    of existing edges.  Each chart rule expects a fixed number of 

    edges, as indicated by the class variable ``NUM_EDGES``.  In 

    particular: 

 

    - A chart rule with ``NUM_EDGES=0`` specifies what new edges are 

      licensed, regardless of existing edges. 

    - A chart rule with ``NUM_EDGES=1`` specifies what new edges are 

      licensed by a single existing edge. 

    - A chart rule with ``NUM_EDGES=2`` specifies what new edges are 

      licensed by a pair of existing edges. 

 

    :type NUM_EDGES: int 

    :cvar NUM_EDGES: The number of existing edges that this rule uses 

        to license new edges.  Typically, this number ranges from zero 

        to two. 

    """ 

    def apply(self, chart, grammar, *edges): 

        """ 

        Add the edges licensed by this rule and the given edges to the 

        chart.  Return a list of the edges that were added. 

 

        :type edges: list(EdgeI) 

        :param edges: A set of existing edges.  The number of edges 

            that should be passed to ``apply`` is specified by the 

            ``NUM_EDGES`` class variable. 

        :rtype: list(EdgeI) 

        """ 

        raise NotImplementedError() 

 

    def apply_iter(self, chart, grammar, *edges): 

        """ 

        Return a generator that will add edges licensed by this rule 

        and the given edges to the chart, one at a time.  Each 

        time the generator is resumed, it will either add a new 

        edge and yield that edge; or return. 

 

        :type edges: list(EdgeI) 

        :param edges: A set of existing edges.  The number of edges 

            that should be passed to ``apply()`` is specified by the 

            ``NUM_EDGES`` class variable. 

        :rtype: iter(EdgeI) 

        """ 

        raise NotImplementedError() 

 

    def apply_everywhere(self, chart, grammar): 

        """ 

        Add all the edges licensed by this rule and the edges in the 

        chart to the chart.  Return a list of the edges that were added. 

 

        :rtype: list(EdgeI) 

        """ 

        raise NotImplementedError() 

 

    def apply_everywhere_iter(self, chart, grammar): 

        """ 

        Return a generator that will add all edges licensed by 

        this rule, given the edges that are currently in the 

        chart, one at a time.  Each time the generator is resumed, 

        it will either add a new edge and yield that edge; or return. 

 

        :rtype: iter(EdgeI) 

        """ 

        raise NotImplementedError() 

 

class AbstractChartRule(ChartRuleI): 

    """ 

    An abstract base class for chart rules.  ``AbstractChartRule`` 

    provides: 

 

    - A default implementation for ``apply``, based on ``apply_iter``. 

    - A default implementation for ``apply_everywhere_iter``, 

      based on ``apply_iter``. 

    - A default implementation for ``apply_everywhere``, based on 

      ``apply_everywhere_iter``.  Currently, this implementation 

      assumes that ``NUM_EDGES``<=3. 

    - A default implementation for ``__str__``, which returns a 

      name basd on the rule's class name. 

    """ 

 

    # Subclasses must define apply_iter. 

    def apply_iter(self, chart, grammar, *edges): 

        raise NotImplementedError() 

 

    # Default: loop through the given number of edges, and call 

    # self.apply() for each set of edges. 

    def apply_everywhere_iter(self, chart, grammar): 

        if self.NUM_EDGES == 0: 

            for new_edge in self.apply_iter(chart, grammar): 

                yield new_edge 

 

        elif self.NUM_EDGES == 1: 

            for e1 in chart: 

                for new_edge in self.apply_iter(chart, grammar, e1): 

                    yield new_edge 

 

        elif self.NUM_EDGES == 2: 

            for e1 in chart: 

                for e2 in chart: 

                    for new_edge in self.apply_iter(chart, grammar, e1, e2): 

                        yield new_edge 

 

        elif self.NUM_EDGES == 3: 

            for e1 in chart: 

                for e2 in chart: 

                    for e3 in chart: 

                        for new_edge in self.apply_iter(chart,grammar,e1,e2,e3): 

                            yield new_edge 

 

        else: 

            raise AssertionError('NUM_EDGES>3 is not currently supported') 

 

    # Default: delegate to apply_iter. 

    def apply(self, chart, grammar, *edges): 

        return list(self.apply_iter(chart, grammar, *edges)) 

 

    # Default: delegate to apply_everywhere_iter. 

    def apply_everywhere(self, chart, grammar): 

        return list(self.apply_everywhere_iter(chart, grammar)) 

 

    # Default: return a name based on the class name. 

    def __str__(self): 

        # Add spaces between InitialCapsWords. 

        return re.sub('([a-z])([A-Z])', r'\1 \2', self.__class__.__name__) 

 

#//////////////////////////////////////////////////////////// 

# Fundamental Rule 

#//////////////////////////////////////////////////////////// 

 

class FundamentalRule(AbstractChartRule): 

    """ 

    A rule that joins two adjacent edges to form a single combined 

    edge.  In particular, this rule specifies that any pair of edges 

 

    - ``[A -> alpha \* B beta][i:j]`` 

    - ``[B -> gamma \*][j:k]`` 

 

    licenses the edge: 

 

    - ``[A -> alpha B * beta][i:j]`` 

    """ 

    NUM_EDGES = 2 

    def apply_iter(self, chart, grammar, left_edge, right_edge): 

        # Make sure the rule is applicable. 

        if not (left_edge.is_incomplete() and 

                right_edge.is_complete() and 

                left_edge.end() == right_edge.start() and 

                next(left_edge) == right_edge.lhs()): 

            return 

 

        # Construct the new edge. 

        new_edge = left_edge.move_dot_forward(right_edge.end()) 

 

        # Insert it into the chart. 

        if chart.insert_with_backpointer(new_edge, left_edge, right_edge): 

            yield new_edge 

 

class SingleEdgeFundamentalRule(FundamentalRule): 

    """ 

    A rule that joins a given edge with adjacent edges in the chart, 

    to form combined edges.  In particular, this rule specifies that 

    either of the edges: 

 

    - ``[A -> alpha \* B beta][i:j]`` 

    - ``[B -> gamma \*][j:k]`` 

 

    licenses the edge: 

 

    - ``[A -> alpha B * beta][i:j]`` 

 

    if the other edge is already in the chart. 

 

    :note: This is basically ``FundamentalRule``, with one edge left 

        unspecified. 

    """ 

    NUM_EDGES = 1 

 

    def apply_iter(self, chart, grammar, edge): 

        if edge.is_incomplete(): 

            for new_edge in self._apply_incomplete(chart, grammar, edge): 

                yield new_edge 

        else: 

            for new_edge in self._apply_complete(chart, grammar, edge): 

                yield new_edge 

 

    def _apply_complete(self, chart, grammar, right_edge): 

        for left_edge in chart.select(end=right_edge.start(), 

                                      is_complete=False, 

                                      next=right_edge.lhs()): 

            new_edge = left_edge.move_dot_forward(right_edge.end()) 

            if chart.insert_with_backpointer(new_edge, left_edge, right_edge): 

                yield new_edge 

 

    def _apply_incomplete(self, chart, grammar, left_edge): 

        for right_edge in chart.select(start=left_edge.end(), 

                                       is_complete=True, 

                                       lhs=next(left_edge)): 

            new_edge = left_edge.move_dot_forward(right_edge.end()) 

            if chart.insert_with_backpointer(new_edge, left_edge, right_edge): 

                yield new_edge 

 

#//////////////////////////////////////////////////////////// 

# Inserting Terminal Leafs 

#//////////////////////////////////////////////////////////// 

 

class LeafInitRule(AbstractChartRule): 

    NUM_EDGES=0 

    def apply_iter(self, chart, grammar): 

        for index in range(chart.num_leaves()): 

            new_edge = LeafEdge(chart.leaf(index), index) 

            if chart.insert(new_edge, ()): 

                yield new_edge 

 

#//////////////////////////////////////////////////////////// 

# Top-Down Prediction 

#//////////////////////////////////////////////////////////// 

 

class TopDownInitRule(AbstractChartRule): 

    """ 

    A rule licensing edges corresponding to the grammar productions for 

    the grammar's start symbol.  In particular, this rule specifies that 

    ``[S -> \* alpha][0:i]`` is licensed for each grammar production 

    ``S -> alpha``, where ``S`` is the grammar's start symbol. 

    """ 

    NUM_EDGES = 0 

    def apply_iter(self, chart, grammar): 

        for prod in grammar.productions(lhs=grammar.start()): 

            new_edge = TreeEdge.from_production(prod, 0) 

            if chart.insert(new_edge, ()): 

                yield new_edge 

 

class TopDownPredictRule(AbstractChartRule): 

    """ 

    A rule licensing edges corresponding to the grammar productions 

    for the nonterminal following an incomplete edge's dot.  In 

    particular, this rule specifies that 

    ``[A -> alpha \* B beta][i:j]`` licenses the edge 

    ``[B -> \* gamma][j:j]`` for each grammar production ``B -> gamma``. 

 

    :note: This rule corresponds to the Predictor Rule in Earley parsing. 

    """ 

    NUM_EDGES = 1 

    def apply_iter(self, chart, grammar, edge): 

        if edge.is_complete(): return 

        for prod in grammar.productions(lhs=next(edge)): 

            new_edge = TreeEdge.from_production(prod, edge.end()) 

            if chart.insert(new_edge, ()): 

                yield new_edge 

 

class CachedTopDownPredictRule(TopDownPredictRule): 

    """ 

    A cached version of ``TopDownPredictRule``.  After the first time 

    this rule is applied to an edge with a given ``end`` and ``next``, 

    it will not generate any more edges for edges with that ``end`` and 

    ``next``. 

 

    If ``chart`` or ``grammar`` are changed, then the cache is flushed. 

    """ 

    def __init__(self): 

        TopDownPredictRule.__init__(self) 

        self._done = {} 

 

    def apply_iter(self, chart, grammar, edge): 

        if edge.is_complete(): return 

        next_edge, index = next(edge), edge.end() 

        if not is_nonterminal(next_edge): return 

 

        # If we've already applied this rule to an edge with the same 

        # next & end, and the chart & grammar have not changed, then 

        # just return (no new edges to add). 

        done = self._done.get((next_edge, index), (None,None)) 

        if done[0] is chart and done[1] is grammar: return 

 

        # Add all the edges indicated by the top down expand rule. 

        for prod in grammar.productions(lhs=next_edge): 

            # If the left corner in the predicted production is 

            # leaf, it must match with the input. 

            if prod.rhs(): 

                first = prod.rhs()[0] 

                if is_terminal(first): 

                    if index >= chart.num_leaves() or first != chart.leaf(index): continue 

 

            new_edge = TreeEdge.from_production(prod, index) 

            if chart.insert(new_edge, ()): 

                yield new_edge 

 

        # Record the fact that we've applied this rule. 

        self._done[next_edge, index] = (chart, grammar) 

 

#//////////////////////////////////////////////////////////// 

# Bottom-Up Prediction 

#//////////////////////////////////////////////////////////// 

 

class BottomUpPredictRule(AbstractChartRule): 

    """ 

    A rule licensing any edge corresponding to a production whose 

    right-hand side begins with a complete edge's left-hand side.  In 

    particular, this rule specifies that ``[A -> alpha \*]`` licenses 

    the edge ``[B -> \* A beta]`` for each grammar production ``B -> A beta``. 

    """ 

    NUM_EDGES = 1 

    def apply_iter(self, chart, grammar, edge): 

        if edge.is_incomplete(): return 

        for prod in grammar.productions(rhs=edge.lhs()): 

            new_edge = TreeEdge.from_production(prod, edge.start()) 

            if chart.insert(new_edge, ()): 

                yield new_edge 

 

class BottomUpPredictCombineRule(BottomUpPredictRule): 

    """ 

    A rule licensing any edge corresponding to a production whose 

    right-hand side begins with a complete edge's left-hand side.  In 

    particular, this rule specifies that ``[A -> alpha \*]`` 

    licenses the edge ``[B -> A \* beta]`` for each grammar 

    production ``B -> A beta``. 

 

    :note: This is like ``BottomUpPredictRule``, but it also applies 

        the ``FundamentalRule`` to the resulting edge. 

    """ 

    NUM_EDGES = 1 

    def apply_iter(self, chart, grammar, edge): 

        if edge.is_incomplete(): return 

        for prod in grammar.productions(rhs=edge.lhs()): 

            new_edge = TreeEdge(edge.span(), prod.lhs(), prod.rhs(), 1) 

            if chart.insert(new_edge, (edge,)): 

                yield new_edge 

 

class EmptyPredictRule(AbstractChartRule): 

    """ 

    A rule that inserts all empty productions as passive edges, 

    in every position in the chart. 

    """ 

    NUM_EDGES = 0 

    def apply_iter(self, chart, grammar): 

        for prod in grammar.productions(empty=True): 

            for index in compat.xrange(chart.num_leaves() + 1): 

                new_edge = TreeEdge.from_production(prod, index) 

                if chart.insert(new_edge, ()): 

                    yield new_edge 

 

 

######################################################################## 

##  Filtered Bottom Up 

######################################################################## 

 

class FilteredSingleEdgeFundamentalRule(SingleEdgeFundamentalRule): 

    def _apply_complete(self, chart, grammar, right_edge): 

        end = right_edge.end() 

        nexttoken = end < chart.num_leaves() and chart.leaf(end) 

        for left_edge in chart.select(end=right_edge.start(), 

                                      is_complete=False, 

                                      next=right_edge.lhs()): 

            if _bottomup_filter(grammar, nexttoken, left_edge.rhs(), left_edge.dot()): 

                new_edge = left_edge.move_dot_forward(right_edge.end()) 

                if chart.insert_with_backpointer(new_edge, left_edge, right_edge): 

                    yield new_edge 

 

    def _apply_incomplete(self, chart, grammar, left_edge): 

        for right_edge in chart.select(start=left_edge.end(), 

                                       is_complete=True, 

                                       lhs=next(left_edge)): 

            end = right_edge.end() 

            nexttoken = end < chart.num_leaves() and chart.leaf(end) 

            if _bottomup_filter(grammar, nexttoken, left_edge.rhs(), left_edge.dot()): 

                new_edge = left_edge.move_dot_forward(right_edge.end()) 

                if chart.insert_with_backpointer(new_edge, left_edge, right_edge): 

                    yield new_edge 

 

class FilteredBottomUpPredictCombineRule(BottomUpPredictCombineRule): 

    def apply_iter(self, chart, grammar, edge): 

        if edge.is_incomplete(): 

            return 

 

        end = edge.end() 

        nexttoken = end < chart.num_leaves() and chart.leaf(end) 

        for prod in grammar.productions(rhs=edge.lhs()): 

            if _bottomup_filter(grammar, nexttoken, prod.rhs()): 

                new_edge = TreeEdge(edge.span(), prod.lhs(), prod.rhs(), 1) 

                if chart.insert(new_edge, (edge,)): 

                    yield new_edge 

 

def _bottomup_filter(grammar, nexttoken, rhs, dot=0): 

    if len(rhs) <= dot + 1: 

        return True 

    _next = rhs[dot + 1] 

    if is_terminal(_next): 

        return nexttoken == _next 

    else: 

        return grammar.is_leftcorner(_next, nexttoken) 

 

 

######################################################################## 

##  Generic Chart Parser 

######################################################################## 

 

TD_STRATEGY = [LeafInitRule(), 

               TopDownInitRule(), 

               CachedTopDownPredictRule(), 

               SingleEdgeFundamentalRule()] 

BU_STRATEGY = [LeafInitRule(), 

               EmptyPredictRule(), 

               BottomUpPredictRule(), 

               SingleEdgeFundamentalRule()] 

BU_LC_STRATEGY = [LeafInitRule(), 

                  EmptyPredictRule(), 

                  BottomUpPredictCombineRule(), 

                  SingleEdgeFundamentalRule()] 

 

LC_STRATEGY = [LeafInitRule(), 

               FilteredBottomUpPredictCombineRule(), 

               FilteredSingleEdgeFundamentalRule()] 

 

class ChartParser(ParserI): 

    """ 

    A generic chart parser.  A "strategy", or list of 

    ``ChartRuleI`` instances, is used to decide what edges to add to 

    the chart.  In particular, ``ChartParser`` uses the following 

    algorithm to parse texts: 

 

    | Until no new edges are added: 

    |   For each *rule* in *strategy*: 

    |     Apply *rule* to any applicable edges in the chart. 

    | Return any complete parses in the chart 

    """ 

    def __init__(self, grammar, strategy=BU_LC_STRATEGY, trace=0, 

                 trace_chart_width=50, use_agenda=True, chart_class=Chart): 

        """ 

        Create a new chart parser, that uses ``grammar`` to parse 

        texts. 

 

        :type grammar: ContextFreeGrammar 

        :param grammar: The grammar used to parse texts. 

        :type strategy: list(ChartRuleI) 

        :param strategy: A list of rules that should be used to decide 

            what edges to add to the chart (top-down strategy by default). 

        :type trace: int 

        :param trace: The level of tracing that should be used when 

            parsing a text.  ``0`` will generate no tracing output; 

            and higher numbers will produce more verbose tracing 

            output. 

        :type trace_chart_width: int 

        :param trace_chart_width: The default total width reserved for 

            the chart in trace output.  The remainder of each line will 

            be used to display edges. 

        :type use_agenda: bool 

        :param use_agenda: Use an optimized agenda-based algorithm, 

            if possible. 

        :param chart_class: The class that should be used to create 

            the parse charts. 

        """ 

        self._grammar = grammar 

        self._strategy = strategy 

        self._trace = trace 

        self._trace_chart_width = trace_chart_width 

        # If the strategy only consists of axioms (NUM_EDGES==0) and 

        # inference rules (NUM_EDGES==1), we can use an agenda-based algorithm: 

        self._use_agenda = use_agenda 

        self._chart_class = chart_class 

 

        self._axioms = [] 

        self._inference_rules = [] 

        for rule in strategy: 

            if rule.NUM_EDGES == 0: 

                self._axioms.append(rule) 

            elif rule.NUM_EDGES == 1: 

                self._inference_rules.append(rule) 

            else: 

                self._use_agenda = False 

 

    def grammar(self): 

        return self._grammar 

 

    def _trace_new_edges(self, chart, rule, new_edges, trace, edge_width): 

        if not trace: return 

        should_print_rule_header = trace > 1 

        for edge in new_edges: 

            if should_print_rule_header: 

                print('%s:' % rule) 

                should_print_rule_header = False 

            print(chart.pp_edge(edge, edge_width)) 

 

    def chart_parse(self, tokens, trace=None): 

        """ 

        Return the final parse ``Chart`` from which all possible 

        parse trees can be extracted. 

 

        :param tokens: The sentence to be parsed 

        :type tokens: list(str) 

        :rtype: Chart 

        """ 

        if trace is None: trace = self._trace 

        trace_new_edges = self._trace_new_edges 

 

        tokens = list(tokens) 

        self._grammar.check_coverage(tokens) 

        chart = self._chart_class(tokens) 

        grammar = self._grammar 

 

        # Width, for printing trace edges. 

        trace_edge_width = self._trace_chart_width // (chart.num_leaves() + 1) 

        if trace: print(chart.pp_leaves(trace_edge_width)) 

 

        if self._use_agenda: 

            # Use an agenda-based algorithm. 

            for axiom in self._axioms: 

                new_edges = axiom.apply(chart, grammar) 

                trace_new_edges(chart, axiom, new_edges, trace, trace_edge_width) 

 

            inference_rules = self._inference_rules 

            agenda = chart.edges() 

            # We reverse the initial agenda, since it is a stack 

            # but chart.edges() functions as a queue. 

            agenda.reverse() 

            while agenda: 

                edge = agenda.pop() 

                for rule in inference_rules: 

                    new_edges = rule.apply_iter(chart, grammar, edge) 

                    if trace: 

                        new_edges = list(new_edges) 

                        trace_new_edges(chart, rule, new_edges, trace, trace_edge_width) 

                    agenda += new_edges 

 

        else: 

            # Do not use an agenda-based algorithm. 

            edges_added = True 

            while edges_added: 

                edges_added = False 

                for rule in self._strategy: 

                    new_edges = rule.apply_everywhere(chart, grammar) 

                    edges_added = len(new_edges) 

                    trace_new_edges(chart, rule, new_edges, trace, trace_edge_width) 

 

        # Return the final chart. 

        return chart 

 

    def nbest_parse(self, tokens, n=None, tree_class=Tree): 

        chart = self.chart_parse(tokens) 

        # Return a list of complete parses. 

        return chart.parses(self._grammar.start(), tree_class=tree_class)[:n] 

 

class TopDownChartParser(ChartParser): 

    """ 

    A ``ChartParser`` using a top-down parsing strategy. 

    See ``ChartParser`` for more information. 

    """ 

    def __init__(self, grammar, **parser_args): 

        ChartParser.__init__(self, grammar, TD_STRATEGY, **parser_args) 

 

class BottomUpChartParser(ChartParser): 

    """ 

    A ``ChartParser`` using a bottom-up parsing strategy. 

    See ``ChartParser`` for more information. 

    """ 

    def __init__(self, grammar, **parser_args): 

        if isinstance(grammar, WeightedGrammar): 

            warnings.warn("BottomUpChartParser only works for ContextFreeGrammar, " 

                          "use BottomUpProbabilisticChartParser instead", 

                          category=DeprecationWarning) 

        ChartParser.__init__(self, grammar, BU_STRATEGY, **parser_args) 

 

class BottomUpLeftCornerChartParser(ChartParser): 

    """ 

    A ``ChartParser`` using a bottom-up left-corner parsing strategy. 

    This strategy is often more efficient than standard bottom-up. 

    See ``ChartParser`` for more information. 

    """ 

    def __init__(self, grammar, **parser_args): 

        ChartParser.__init__(self, grammar, BU_LC_STRATEGY, **parser_args) 

 

class LeftCornerChartParser(ChartParser): 

    def __init__(self, grammar, **parser_args): 

        if not grammar.is_nonempty(): 

            raise ValueError("LeftCornerParser only works for grammars " 

                             "without empty productions.") 

        ChartParser.__init__(self, grammar, LC_STRATEGY, **parser_args) 

 

######################################################################## 

##  Stepping Chart Parser 

######################################################################## 

 

class SteppingChartParser(ChartParser): 

    """ 

    A ``ChartParser`` that allows you to step through the parsing 

    process, adding a single edge at a time.  It also allows you to 

    change the parser's strategy or grammar midway through parsing a 

    text. 

 

    The ``initialize`` method is used to start parsing a text.  ``step`` 

    adds a single edge to the chart.  ``set_strategy`` changes the 

    strategy used by the chart parser.  ``parses`` returns the set of 

    parses that has been found by the chart parser. 

 

    :ivar _restart: Records whether the parser's strategy, grammar, 

        or chart has been changed.  If so, then ``step`` must restart 

        the parsing algorithm. 

    """ 

    def __init__(self, grammar, strategy=[], trace=0): 

        self._chart = None 

        self._current_chartrule = None 

        self._restart = False 

        ChartParser.__init__(self, grammar, strategy, trace) 

 

    #//////////////////////////////////////////////////////////// 

    # Initialization 

    #//////////////////////////////////////////////////////////// 

 

    def initialize(self, tokens): 

        "Begin parsing the given tokens." 

        self._chart = Chart(list(tokens)) 

        self._restart = True 

 

    #//////////////////////////////////////////////////////////// 

    # Stepping 

    #//////////////////////////////////////////////////////////// 

 

    def step(self): 

        """ 

        Return a generator that adds edges to the chart, one at a 

        time.  Each time the generator is resumed, it adds a single 

        edge and yields that edge.  If no more edges can be added, 

        then it yields None. 

 

        If the parser's strategy, grammar, or chart is changed, then 

        the generator will continue adding edges using the new 

        strategy, grammar, or chart. 

 

        Note that this generator never terminates, since the grammar 

        or strategy might be changed to values that would add new 

        edges.  Instead, it yields None when no more edges can be 

        added with the current strategy and grammar. 

        """ 

        if self._chart is None: 

            raise ValueError('Parser must be initialized first') 

        while True: 

            self._restart = False 

            w = 50 // (self._chart.num_leaves()+1) 

 

            for e in self._parse(): 

                if self._trace > 1: print(self._current_chartrule) 

                if self._trace > 0: print(self._chart.pp_edge(e,w)) 

                yield e 

                if self._restart: break 

            else: 

                yield None # No more edges. 

 

    def _parse(self): 

        """ 

        A generator that implements the actual parsing algorithm. 

        ``step`` iterates through this generator, and restarts it 

        whenever the parser's strategy, grammar, or chart is modified. 

        """ 

        chart = self._chart 

        grammar = self._grammar 

        edges_added = 1 

        while edges_added > 0: 

            edges_added = 0 

            for rule in self._strategy: 

                self._current_chartrule = rule 

                for e in rule.apply_everywhere_iter(chart, grammar): 

                    edges_added += 1 

                    yield e 

 

    #//////////////////////////////////////////////////////////// 

    # Accessors 

    #//////////////////////////////////////////////////////////// 

 

    def strategy(self): 

        "Return the strategy used by this parser." 

        return self._strategy 

 

    def grammar(self): 

        "Return the grammar used by this parser." 

        return self._grammar 

 

    def chart(self): 

        "Return the chart that is used by this parser." 

        return self._chart 

 

    def current_chartrule(self): 

        "Return the chart rule used to generate the most recent edge." 

        return self._current_chartrule 

 

    def parses(self, tree_class=Tree): 

        "Return the parse trees currently contained in the chart." 

        return self._chart.parses(self._grammar.start(), tree_class) 

 

    #//////////////////////////////////////////////////////////// 

    # Parser modification 

    #//////////////////////////////////////////////////////////// 

 

    def set_strategy(self, strategy): 

        """ 

        Change the strategy that the parser uses to decide which edges 

        to add to the chart. 

 

        :type strategy: list(ChartRuleI) 

        :param strategy: A list of rules that should be used to decide 

            what edges to add to the chart. 

        """ 

        if strategy == self._strategy: return 

        self._strategy = strategy[:] # Make a copy. 

        self._restart = True 

 

    def set_grammar(self, grammar): 

        "Change the grammar used by the parser." 

        if grammar is self._grammar: return 

        self._grammar = grammar 

        self._restart = True 

 

    def set_chart(self, chart): 

        "Load a given chart into the chart parser." 

        if chart is self._chart: return 

        self._chart = chart 

        self._restart = True 

 

    #//////////////////////////////////////////////////////////// 

    # Standard parser methods 

    #//////////////////////////////////////////////////////////// 

 

    def nbest_parse(self, tokens, n=None, tree_class=Tree): 

        tokens = list(tokens) 

        self._grammar.check_coverage(tokens) 

 

        # Initialize ourselves. 

        self.initialize(tokens) 

 

        # Step until no more edges are generated. 

        for e in self.step(): 

            if e is None: break 

 

        # Return a list of complete parses. 

        return self.parses(tree_class=tree_class)[:n] 

 

######################################################################## 

##  Demo Code 

######################################################################## 

 

def demo_grammar(): 

    from nltk.grammar import parse_cfg 

    return parse_cfg(""" 

S  -> NP VP 

PP -> "with" NP 

NP -> NP PP 

VP -> VP PP 

VP -> Verb NP 

VP -> Verb 

NP -> Det Noun 

NP -> "John" 

NP -> "I" 

Det -> "the" 

Det -> "my" 

Det -> "a" 

Noun -> "dog" 

Noun -> "cookie" 

Verb -> "ate" 

Verb -> "saw" 

Prep -> "with" 

Prep -> "under" 

""") 

 

def demo(choice=None, 

         should_print_times=True, should_print_grammar=False, 

         should_print_trees=True, trace=2, 

         sent='I saw John with a dog with my cookie', numparses=5): 

    """ 

    A demonstration of the chart parsers. 

    """ 

    import sys, time 

    from nltk import nonterminals, Production, ContextFreeGrammar 

 

    # The grammar for ChartParser and SteppingChartParser: 

    grammar = demo_grammar() 

    if should_print_grammar: 

        print("* Grammar") 

        print(grammar) 

 

    # Tokenize the sample sentence. 

    print("* Sentence:") 

    print(sent) 

    tokens = sent.split() 

    print(tokens) 

    print() 

 

    # Ask the user which parser to test, 

    # if the parser wasn't provided as an argument 

    if choice is None: 

        print('  1: Top-down chart parser') 

        print('  2: Bottom-up chart parser') 

        print('  3: Bottom-up left-corner chart parser') 

        print('  4: Left-corner chart parser with bottom-up filter') 

        print('  5: Stepping chart parser (alternating top-down & bottom-up)') 

        print('  6: All parsers') 

        print('\nWhich parser (1-6)? ', end=' ') 

        choice = sys.stdin.readline().strip() 

        print() 

 

    choice = str(choice) 

    if choice not in "123456": 

        print('Bad parser number') 

        return 

 

    # Keep track of how long each parser takes. 

    times = {} 

 

    strategies = {'1': ('Top-down', TD_STRATEGY), 

                  '2': ('Bottom-up', BU_STRATEGY), 

                  '3': ('Bottom-up left-corner', BU_LC_STRATEGY), 

                  '4': ('Filtered left-corner', LC_STRATEGY)} 

    choices = [] 

    if choice in strategies: choices = [choice] 

    if choice=='6': choices = "1234" 

 

    # Run the requested chart parser(s), except the stepping parser. 

    for strategy in choices: 

        print("* Strategy: " + strategies[strategy][0]) 

        print() 

        cp = ChartParser(grammar, strategies[strategy][1], trace=trace) 

        t = time.time() 

        chart = cp.chart_parse(tokens) 

        parses = chart.parses(grammar.start()) 

        times[strategies[strategy][0]] = time.time()-t 

        print("Nr edges in chart:", len(chart.edges())) 

        if numparses: 

            assert len(parses)==numparses, 'Not all parses found' 

        if should_print_trees: 

            for tree in parses: print(tree) 

        else: 

            print("Nr trees:", len(parses)) 

        print() 

 

    # Run the stepping parser, if requested. 

    if choice in "56": 

        print("* Strategy: Stepping (top-down vs bottom-up)") 

        print() 

        t = time.time() 

        cp = SteppingChartParser(grammar, trace=trace) 

        cp.initialize(tokens) 

        for i in range(5): 

            print('*** SWITCH TO TOP DOWN') 

            cp.set_strategy(TD_STRATEGY) 

            for j, e in enumerate(cp.step()): 

                if j>20 or e is None: break 

            print('*** SWITCH TO BOTTOM UP') 

            cp.set_strategy(BU_STRATEGY) 

            for j, e in enumerate(cp.step()): 

                if j>20 or e is None: break 

        times['Stepping'] = time.time()-t 

        print("Nr edges in chart:", len(cp.chart().edges())) 

        if numparses: 

            assert len(cp.parses())==numparses, 'Not all parses found' 

        if should_print_trees: 

            for tree in cp.parses(): print(tree) 

        else: 

            print("Nr trees:", len(cp.parses())) 

        print() 

 

    # Print the times of all parsers: 

    if not (should_print_times and times): return 

    print("* Parsing times") 

    print() 

    maxlen = max(len(key) for key in times) 

    format = '%' + repr(maxlen) + 's parser: %6.3fsec' 

    times_items = times.items() 

    times_items.sort(lambda a,b:cmp(a[1], b[1])) 

    for (parser, t) in times_items: 

        print(format % (parser, t)) 

 

if __name__ == '__main__': demo()