Proactive Spoken Dialogue Interaction in Multi-Party Environments (eBook)

Preface 5
Contents 6
List of Figures 9
List of Tables 11
1 Introduction 12
1.1 Introduction on Spoken Language Dialogue Systems 13
1.1.1 System Architecture 13
1.1.2 Current Trends in Spoken Language Dialogue Systems 17
1.2 Related Work on Advanced Dialogue Systems 19
1.3 The Computer as a Dialogue Partner 20
1.4 Challenges 24
1.5 Outline of the Book 25
2 Fundamentals 27
2.1 Corpus Development 27
2.2 Evaluation of Spoken Language Dialolgue Systems 30
2.3 Multi-Party Interaction 32
2.3.1 Speech Acts and other Linguistic Fundamentals 32
2.3.2 Conversational Roles 35
2.3.3 Human-Human and Human-Computer Interaction 38
2.4 Dialogue Modelling 44
2.4.1 Dialogue Context and History 45
2.4.2 Dialogue Management 47
2.4.3 Information State Update Approach To Dialogue Modelling 49
2.4.4 Multi-Party Dialogue Modelling 53
2.5 Summary 59
3 Multi-Party Dialogue Corpus 60
3.1 Existing Multi-Party Corpora 60
3.2 Wizard-of-Oz Data Collection 64
3.2.1 Experimental Setup 64
3.2.2 Procedure 65
3.2.3 System Interaction Policies 67
3.2.4 WIT: The Wizard Interaction Tool 69
3.3 The PIT Corpus 74
3.3.1 Data Structure 75
3.3.2 Annotation 76
3.3.3 Dialogue Analysis 77
3.4 Summary 81
4 Dialogue Management for a Multi-Party Spoken Dialogue System 82
4.1 Multi-Party Dialogue Modelling 84
4.1.1 Dialogue Model 84
4.1.2 Interaction Protocols 87
4.2 Dialogue Management in the Example Domain of Restaurant Selection 90
4.2.1 Dialogue Context 90
4.2.2 Domain Model 90
4.2.3 Task Model 91
4.2.4 Information State Updates 93
4.2.5 Dialogue Plans 100
4.3 Enabling Proactiveness 101
4.3.1 Optimistic Grounding and Integration Strategy for Multi-Party Setup 101
4.3.2 System Interaction Strategy 103
4.3.3 Dialogue History for Proactive System Interaction 107
4.4 Proactive Dialogue Management Example 111
4.5 Problem Solving Using Discourse Motivated Constraint Prioritisation 116
4.5.1 Prioritisation Scheme 118
4.5.2 Example 121
4.6 Summary 122
5 Evaluation 124
5.1 Usability Evaluation 125
5.1.1 Questionnaire Design 125
5.1.2 Participants 127
5.1.3 Analysing the System Progress 129
5.1.4 Assessing the Usability 131
5.2 Evaluating System Performance 133
5.2.1 Descriptive Analysis of the PIT Corpus 134
5.2.2 Evaluation of Discourse Motivated Constraint Prioritisation 136
5.3 Gaze Direction Analysis to Assess User Acceptance 138
5.4 Assessing Proactiveness 142
5.4.1 Addressing Behaviour During First Interaction Request 142
5.4.2 E ect of Avatar on Proactiveness 143
5.4.3 Subjective Evaluation 145
5.5 Summary 147
6 Conclusions and Future Directions 149
6.1 Summary 149
6.2 Future Directions 154
A Wizard Interaction Tool 158
B Example Dialogue 162
C Questionnaire 164
Index 168
References 171