Learning Machine Translation
Cyril Goutte
Nicola Cancedda
Marc Dymetman
George Foster
The MIT Press
Contents
Series Foreword ...................................................................................................... ix
Preface............................................................................................................................. xi
1 A Statistical Machine Translation Primer.............................................................. 1
Nicola Cancedda, Marc Dymetman, George Foster, and Cyril Goutte
1.1
Background.............................................................................................................. 1
1.2
Evaluation
of Machine Translation..................................................................... 3
1.3
Word-Based
MT...................................................................................................... 8
1.4
Language
Models.................................................................................................... 11
1.5
Phrase-Based
MT .......................................................................................... 18
1.6
Syntax-Based
SMT................................................................................................. 26
1.7
Some
Other Important Directions.............................................................. 30
1.8
Machine
Learning for SMT ............................................................................... 32
1.9
Conclusion ............................................................................................................ 36
1 Enabling Technologies 39
2 Mining Patents for Parallel Corpora............................................................................... 41
Masao Utiyama and Hitoshi Isahara
2.1
Introduction.......................................................................................................... 41
2.2
Related
Work........................................................................................................... 42
2.3
Resources.................................................................................................................. 44
2.4
Alignment
Procedure............................................................................................... 44
2.5
Statistics
of the Patent Parallel Corpus.................................................. 48
2.6
MT
Experiments................................................................................................. 51
2.7
Conclusion ............................................................................................................ 56
3 Automatic Construction of Multilingual Name Dictionaries . . 59
Bruno Pouliquen and Ralf Steinberger
3.1
Introduction
and Motivation....................................................................... 59
3.2
Related
Work........................................................................................................... 65
3.3
Multilingual
Recognition of New Names............................................................ 68
3.4
Lookup
of Known Names and Their Morphological Variants....................... 70
3.5
Evaluation
of Person Name Recognition…………………………………… 72
3.6
Identification
and Merging of Name Variants …………………………… 74
3.7
Conclusion
and Future Work ………………………………………………
78
4... Named Entity
Transliteration and Discovery in Multilingual Cor
pora 79
Alexandre Klementiev and Dan Roth
4.1
Introduction …………………………………………………………79
4.2
Previous
Work ………………………………………………………….82
4.3
Co-Ranking:
An Algorithm for NE
Discovery………………………………..83
4.4
Experimental
Study
…………………………………………………………86
4.5
Conclusions …………………………………………………………91
4.6
Future
Work …………………………………………………………92
5... Combination of
Statistical Word Alignments Based on Multiple
Preprocessing Schemes …………………………………………………………..93
Jakob Elming, Nizar Habash, and Josep M. Crego
5.1
Introduction …………………………………………………………93
5.2
Related
Work …………………………………………………………94
5.3
Arabic
Preprocessing Schemes ……………………………………………….95
5.4
Preprocessing Schemes for Alignment ……………………………………….96
5.5
Alignment
Combination …………………………………………………….. 97
5.6
Evaluation …………………………………………………………99
5.7
Postface: Machine Translation and Alignment Improvements …………..107
5.8
Conclusion ………………………………………………………..110
6 Linguistically Enriched
Word-Sequence Kernels for Discriminative
Language Modeling …………………………………………………….. ….. 111
Pierre Mahé and Nicola Cancedda
6.1
Motivations
111
6.2
Linguistically
Enriched Word-Sequence Kernels ……………………………113
6.3
Experimental
Validation ……………………………………………………..119
6.4
Conclusion
and Future Work …………………………………………………125
II Machine Translation ……………………………………………………….. 129
7 Toward Purely Discriminative Training for
Tree-Structured Trans
lation Models ……………………………………………………………. 131
Benjamin
Wellington, Joseph Turian, and
7.1
Introduction …………………………………………………………131
7.2
Related
Work …………………………………………………………132
7.3
Learning
Method …………………………………………………………134
7.4
Experiments …………………………………………………………140
7.5
Conclusion …………………………………………………………148
8 Reranking for Large-Scale Statistical Machine Translation ………………...151
Kenji Yamada and Ion Muslea
8.1
Introduction ………………………………………………………….151
8.2
Background ………………………………………………………….152
8.3
Related
Work ………………………………………………………….153
8.4
Our
Approach ………………………………………………………….154
8.5
Experiment
1: Reranking for the Chinese-to-English System ……………...156
8.6
Experiment
2: Reranking for the French-to-English System ………………..161
8.7
Discussion …………………………………………………………..165
8.8
Conclusion …………………………………………………………..165
9 Kernel-Based Machine Translation ………………………………………………169
Zhuoran Wang and John Shawe-Taylor
9.1
Introduction ..…………………………………………………………169
9.2
Regression
Modeling for SMT ………………………………………………..171
9.3
Decoding …………………………………………………………...175
9.4
Experiments ……………………………………………………………177
9.5
Further
Discussions ……………………………………………………….........182
9.6
Conclusion ……………………………………………………………183
10.. Statistical
Machine Translation through Global Lexical Selection
and Sentence Reconstruction ………………………………………………….. 185
Srinivas Bangalore, Stephan Kanthak, and Patrick Haffner
10.1
Introduction ……………………………………………………… 185
10.2
SFST
Training and Decoding ……………………………………………….187
10.3
Discriminant Models for Lexical Selection …………………………………193
10.4
Choosing
the Classifier ……………………………………………………….195
10.5
Data
and Experiments …………………………………………………..198
10.6
Discussion ………………………………………………………….201
10.7
Conclusions ………………………………………………………….202
11 Discriminative Phrase Selection for SMT ………………………………………205
Jesús Giménez and Lluís Mŕrquez
11.1
Introduction ………………………………………………………….205
11.2
Approaches
to Dedicated Word Selection ………………………………….207
11.3
Discriminative
Phrase Translation ………………………………………,,.209
11.4
Local
Phrase Translation …………………………………………212
11.5
Exploiting
Local DPT Models for the Global Task ………………218
11.6
Conclusions ………………………………………………………….234
12 Semisupervised Learning for Machine Translation …………………………….237
Nicola Ueffing, Gholamreza Haffari, and Anoop Sarkar
12.1
Introduction …………………………………………………………….237
12.2
12.3
The
Framework …………………………………………………………….240
12.4
Experimental
Results …………………………………………………………..245
12.5
Previous
Work …………………………………………………………….253
12.6
Conclusion
and Outlook ……………………………………………………….255
13 Learning to Combine Machine Translation Systems …………………………….257
Evgeny Matusov, Gregor Leusch, and Hermann Ney
13.1
Introduction …………………………………………………………….257
13.2
Word
Alignment …………………………………………………………….260
13.3
Confusion
Network Generation and Scoring ………………………………….266
13.4
Experiments …………………………………………………………….272
13.5
Conclusion …………………………………………………………….276
References …………………………………………………………………………...277
Contributors ……………………………………………………………………….307
Index ………………………………………………………………………………….313