口语机器翻译2025|PDF|Epub|mobi|kindle电子书版本百度云盘下载

口语机器翻译PDF格式电子书版下载

注意：本站所有压缩包均有解压码： 点击下载压缩包解压工具

1 Introduction1

1.1　What This Book Is About1

1.1.1　Why Do Spoken Language Translation?2

1.1.2　What Are the Basic Problems?2

1.1.3　What Is It Realistic to Attempt Today?4

1.1.4　What Have We Achieved?5

1.2　Overall System Architecture6

1.3　An Illustrative Example9

1.4　In Defence of Hand-Coded Grammars12

1.5　Hybrid Transfer16

1.5.1　The Need for Grammatical Knowledge16

1.5.2　The Need for Preferences17

1.6　Speech Processing20

1.7 Corpora21

Part Ⅰ　Language Processing and Corpora25

2　Translation Using the Core Language Engine25

2.1 Introduction:Multi-Engine Translation25

2.2　Word-to-Word Translation26

2.3　Quasi Logical Form27

2.3.1 Introduction27

2.3.2　Structure of QLF28

2.3.3 QLF as a Transfer Formalism:Examples32

2.3.4　Head-Head Relations in QLF33

2.4　Unification Grammar and QLFs35

2.4.1　The CLE Unification Grammar Formalism35

2.4.2 Unification Grammar Example:French Noun Phrases37

2.4.3　Example 2a:Clauses in Swedish41

2.4.4 Example 2b:Relative Clauses in Swedish42

2.5　Orthographic Analysis and the Lexicon45

2.6　Transfer Rules48

2.6.1　Pre-and Posttransfer50

2.7　The QLF-Based Processing Path51

2.7.1　Linguistic Analysis51

2.7.2　Transfer and Transfer Preferences53

2.7.3 Generation55

2.8　Summary55

3 Grammar Specialisation57

3.1 Introduction57

3.2 Explanation-Based Learning for Grammar Specialisation58

3.2.1 A Definition of Explanation-Based Learning58

3.2.2 Explanation-Based Learning on Unification Grammars61

3.2.3　Category Specialisation62

3.2.4　Elaborate Cutting-Up Criteria65

3.3　An LR Parsing Method for Specialised Grammars67

3.3.1　Basic LR Parsing67

3.3.2　Prefix Merging67

3.3.3　Abstraction68

3.4　Empirical Results69

3.4.1　Experimental Setup69

3.4.2　Discussion of Results71

3.5　Conclusions77

4　Choosing among Interpretations78

4.1　Properties and Discriminants78

4.2　Constituent Pruning82

4.2.1　Discriminants for Pruning82

4.2.2　Deciding Which Edges to Prune85

4.2.3　Probability Estimates for Discriminants85

4.2.4　Relation to Other Pruning Methods89

4.3　Choosing among QLF Analyses90

4.3.1　Analysis Choice:An Example90

4.3.2 Further Advantages of a Discriminant Scheme91

4.3.3 Numerical Metrics92

4.4　Choosing among Transferred QLFs94

4.5　Choosing Paths in the Chart95

5　The TreeBanker98

5.1　Motivation98

5.2 Representational Issues99

5.3　Overview of the TreeBanker100

5.4　The Supervised Training Process100

5.4.1　Properties and Discriminants in Training101

5.4.2 Additional Functionality105

5.5　Training for Transfer Choice106

5.6 Evaluation and Conclusions108

6　Acquisition of Lexical Entries110

6.1　The Lexical Acquisition Tool,LexMake110

6.2 Acquiring Word-to-Word Transfer Rules114

6.3　Evaluation and Conclusions115

7　Spelling and Morphology116

7.1 Introduction116

7.2　The Description Language118

7.2.1　Morphophonology118

7.2.2 Word Formation and Interfacing to Syntax120

7.3 Compilation121

7.3.1　Compiling Spelling Patterns121

7.3.2 Representing Lexical Roots122

7.3.3　Applying Obligatory Rules123

7.3.4　Interword Rules124

7.3.5　Timings124

7.4　Some Examples125

7.4.1 Multiple-Letter Spelling Changes125

7.4.2 Using Features to Control Rule Application126

7.4.3　Interword Spelling Changes127

7.5　Debugging the Rules128

7.6 Conclusions and Further Work130

8　Corpora and Data Collection131

8.1 Rationale and Requirements131

8.2　Simulation Methodology133

8.2.1 Wizard-of-Oz Simulations133

8.2.2　American ATIS Simulations133

8.2.3 Swedish ATIS Simulations134

8.3　Translations of American WOZ Material135

8.3.1　Translations:A First Step135

8.3.2　Email Corpus136

8.4　A Comparison of the Corpora137

8.5 Concluding Remarks on Corpus Collection139

8.6　Representative Corpora and Rational Development141

Part Ⅱ　Linguistic Coverage147

9　English Coverage147

9.1 Overview of English Linguistic Coverage147

9.2 Lexical Items148

9.3　The English Grammar148

9.3.1 Noun Phrases148

9.3.2 Nonrecursive NPs149

9.3.3 Recursive NPs153

9.3.4 Prepositional Phrases156

9.3.5 Numbers157

9.3.6　Verb Phrases157

9.3.7 Clauses and Top-Level Utterances161

9.4　Coverage Failures164

9.5　Comparison with French and Swedish Grammars166

10　French Coverage168

10.1　Introduction168

10.2　Question Formation169

10.2.1 Constraints on Question Formation169

10.2.2　Implementation of the Rules for Question Formation174

10.2.3　Empirical Evaluation Using a Multidimensional Test Suite175

10.3　Clitics176

10.4　Agreement178

10.5　Conclusions179

11　Swedish Coverage180

11.1　Introduction180

11.2　Clausal Constructions181

11.2.1 Inverted Word Order and Verb-Second Phenomena181

11.2.2　Adverbs and Negation183

11.2.3　Other Clausal Constructions185

11.3　Verbs and Verbal Constructions186

11.4　NP Constructions188

11.4.1　Compound Nominals190

12　Transfer Coverage192

12.1　Introduction192

12.2　Statistical Breakdown of Rule Types193

12.3　Overview of the Rules194

12.3.1　Identity Rules194

12.3.2　Lexical Rules Translating Atoms into Atoms195

12.3.3　Lexical Rules Translating Nonatomic Fixed Structures196

12.3.4　Date,Time,and Code Expressions197

12.3.5 Nominals198

12.3.6 Verbs199

12.3.7 Adjectives201

12.3.8　Prepositional Phrases201

12.3.9　Tense,Aspect,Mood,and Voice202

12.3.10　Determiners and Pronouns203

12.3.11　Conjunction205

12.4　Adequacy of the Transfer Formalism206

12.4.1 Expressiveness of the Rule Formalism207

12.4.2　Formal Properties210

12.5 Summary211

13　Rational Reuse of Linguistic Data212

13.1 Introduction212

13.2　Porting Grammars and Lexica among Closely Related Languages213

13.3　Transfer Composition216

13.3.1 Introduction216

13.3.2　Transfer Composition as a Program Transformation217

13.3.3　Procedural Realisation of Transfer Rule Composition220

13.3.4　Composing Transfer Preferences222

13.3.5　Improving Automatically Composed Rule Sets222

13.4　Experiments223

13.4.1　Swedish→English→French224

13.4.2 English→Swedish→Danish226

13.5　Evaluation227

13.6　Conclusions228

Part Ⅲ　Speech Processing231

14　Speech Recognition231

14.1　Speech Recognition Based on Statistical Methods231

14.2　Hidden Markov Models232

14.2.1　Definition232

14.2.2　Observation-Probability Computation234

14.2.3 Estimation of the Hidden-State Sequence236

14.2.4 Estimation of Model Parameters237

14.3　The Speech Part of the Book239

15　Acoustic Modelling240

15.1 Introduction:Discrete or Continuous?That's the Question240

15.2　Continuous-Density HMMs and Genones241

15.3　Efficiency Issues244

15.3.1　Baseline Experiments244

15.3.2　Speed Optimisation245

15.4　Discrete-Mixture HMMs247

15.5 Conclusions249

16　Language Modelling for Multilingual Speech Translation250

16.1　Introduction250

16.2　Fabricating Domain-Specific Data251

16.3 Better Use of Domain-General Data253

16.4　Unsupervised Language-Model Adaptation255

16.5　Phrase-Based Language Models256

16.6 Multilingual Language Modelling262

16.7 Conclusions263

17　Porting a Recogniser to a New Language265

17.1　The Swedish Speech Corpus265

17.1.1 Read-Text Corpora265

17.1.2　WOZ Corpora266

17.2　The Swedish Lexicon267

17.2.1　Phone Set267

17.2.2　Phonetic Transcription268

17.2.3　Lexicon Statistics270

17.3　Acoustic Models270

17.3.1 SLT 2 Models270

17.3.2 SLT-3 Models272

17.4 Conclusions273

18　Multiple Dialects and Languages274

18.1 Introduction274

18.2　Dialect Adaptation274

18.2.1　Dialect Adaptation Methods275

18.2.2　Experimental Results277

18.3　The Multilingual Speech-Recognition System280

18.3.1　Multilingual Recognition Experiments281

18.3.2　Language Identification282

18.4　Conclusions283

19 Common Speech/Language Issues284

19.1　The Speech/Language Interface284

19.2　Split verses Unsplit Compounds in Speech Understanding285

19.2.1　Introduction285

19.2.2　Speech Recognition Experiments286

19.2.3　Conclusions289

19.3　Prosody Translation290

19.3.1　Detection of Focal Accent290

19.3.2　Prosody Transfer293

Part Ⅳ　Evaluation and Conclusions297

20　Evaluation297

20.1　Methodological Issues297

20.2　Evaluation of Speech-to-Text Translation298

20.3　Evaluation of Speech-to-Speech Translation299

20.4　Speech-to-Text Evaluation Results302

20.5　Pipeline Synergy309

21 Conclusions313

A Appendix:The Mathematics of Discriminant Scores315

B　Appendix:Notation for QLF-Based Processing318

B.1 QLFs318

B.2　Grammar Rules320

B.3　Lexicon322

References323

Index333