ISBN: 3-540-66753-9
TITLE: Incremental Speech Translation
AUTHOR: Amtrup, Jan W.
TOC:

Overview 1
1. Introduction 3
1.1 Incremental Natural Language Processing 3
1.2 Incremental Speech Understanding 11
1.3 Incremental Architectures and the Architecture of MILC 15
1.4 Summary 24
2. Graph Theory and Natural Language Processing 25
2.1 General Definitions 25
2.2 The Use of Word Graphs for Natural Language Processing Systems 30
2.3 Evaluation of Word Graphs: Size and Quality Measures 34
2.4 Evaluation of Word Graphs: Quality Measures 44
2.5 Further Operations on Word Graphs 48
2.5.1 Removing Isolated Silence 48
2.5.2 Removing Consecutive Silence 49
2.5.3 Removing All Silence Edges 51
2.5.4 Merging Mutually Unreachable Vertices 51
2.6 Hypergraphs 52
2.6.1 Formal Definition of Hypergraphs 54
2.6.2 Merging of Hyperedges 56
2.6.3 Combination of Hyperedges 59
2.7 Search in Graphs 60
2.8 Summary 62
3. Unification-Based Formalisms for Translation in Natural Language Processing 65
3.1 Unification-Based Formalisms for Natural Language Processing 65
3.1.1 Definition of Typed Feature Structures with Appropriateness 68
3.1.2 Type Lattices 68
3.1.3 Feature Structures 69
3.1.4 Functions as Values of Features 73
3.2 Unification-Based Machine Translation 73
3.3 Architecture and Implementation of the Formalism 76
3.3.1 Definition and Implementation of Type Lattices 79
3.3.2 Definition and Implementation of Feature Structures 80
3.4 Summary 84
4. MILC: Structure and Implementation 85
4.1 Layered Charts 86
4.2 Communication Within the Application 95
4.2.1 Communication Architecture of an Application 96
4.2.2 Channel Models 98
4.2.3 Information Service and Synchronization 100
4.2.4 Termination 104
4.3 Overview of the Architecture of MILC 105
4.4 Word Recognition 106
4.5 Idiom Processing 108
4.6 Parsing 110
4.6.1 Derivation of Verbal Complexes 111
4.6.2 Spontaneous Speech and Word Recognition 113
4.6.3 Structure and Processing Strategies 115
4.7 Utterance Integration 121
4.8 Transfer 128
4.8.1 Chart-Based Transfer 130
4.8.2 The Implementation of Transfer for MILC 132
4.9 Generation 137
4.10 Visualization 143
4.11 Extensions 145
4.11.1 Extension of the Architecture 147
4.11.2 Anytime Translation 149
4.12 System Size 152
4.13 Summary 152
5. Experiments and Results 155
5.1 Hypergraphs 156
5.2 Translation 158
5.2.1 Data Material 158
5.2.2 Linguistic Knowledge Sources 159
5.2.3 Experiments and System Parameters 161
5.2.4 Evaluation 162
5.2.5 Extensions 164
5.3 Comparison With Non-incremental Methods 165
5.4 Summary 167
6. Conclusion and Outlook 169
Bibliography 175
Glossary 193
Index 195
END
