Inhalt

•  
    From HHI to HAI: Developing a Model
•  
   1 Introduction
•  
   1.1 Smart Home Interaction
•  
   1.2 Research Question and Hypothesis
•  
   1.3 Thesis Outline and Contribution
•  
  2 Human-Human Interaction
•  
  2.1 Attention in HHI
•  
   2.1.1 Theory of Attention
•  
   2.1.2 Concepts and Definitions
•  
   2.1.3 The Role of Eye Gaze
•  
  2.2 Hesitations and their Role in HHI
•  
   2.2.1 Disfluent Speech and Hesitation Definition
•  
   2.2.2 Hesitations: Symptom or Signal?
•  
   2.2.3 Effects on the Listener
•  
   2.3 Summary of Research on Attention and Hesitations in HHI
•  
  3 Human-Agent Interaction
•  
  3.1 Attention in Human-Agent Dialogue
•  
   3.1.1 Visual Attention in Interaction
•  
   3.1.2 Attention Models and Measurements
•  
  3.2 Hesitations in Human-Agent Interaction
•  
   3.2.1 Detecting Hesitations
•  
   3.2.2 Hesitations as an Intervention Strategy
•  
   3.3 Summary of Research on Attention and Hesitations in HAI
•  
  4 The Attention-Hesitation Model (AHM)
•  
   4.1 Interaction phases and Disturbances
•  
   4.2 Dialogue Management Responsibilities
•  
   4.2.1 When to (re-)act: From Visual Attention to a Cognitive Model
•  
   4.2.2 How to (re-)act: Hesitation Intervention Strategy
•  
   4.3 Differences to Other Models
•  
   5 Summary of Part I
•  
    Fundamentals for Autonomous HAI
•  
  6 Modeling Dialogue for HAI
•  
   6.1 Architecture of Dialogue Systems
•  
   6.1.1 The Classical Natural Language Processing Pipeline
•  
   6.1.2 Drawbacks of the Conceptual Architecture
•  
   6.1.3 Coordination of Dialogue
•  
   6.1.4 Incremental Dialogue Processing
•  
  6.2 Resulting Requirements: System Engineering Perspective
•  
   6.2.1 Software Requirements
•  
   6.2.2 Hardware Requirements
•  
  7 Realization of the Dialogue System
•  
  7.1 Research Platform
•  
   7.1.1 Cognitive Service Robotics Apartment
•  
   7.1.2 Anthropomorphic Robot Head Flobi
•  
  7.2 General Software Architecture
•  
   7.2.1 Abstraction Layers of the CSRA
•  
   7.2.2 Middleware and Interfaces
•  
   7.2.3 Continuous Integration
•  
   7.3 Dialogue Modules
•  
   7.3.1 Speech Recognition and Understanding
•  
   7.3.2 Decision Management—the Dialogue Manager
•  
   7.3.3 Speech Output and Other Actors
•  
   7.3.4 Introspection Capabilities
•  
  7.4 Meeting the Requirements
•  
   7.4.1 Hardware Requirements
•  
   7.4.2 Software Requirements
•  
  7.5 Dialogue Interaction Scenarios
•  
   7.5.1 Interaction Zones within the CSRA
•  
   7.5.2 Simple Service Robot Interaction Scenario
•  
   7.5.3 Further Interaction Scenarios without Agents
•  
   8 Summary of Part II
•  
    Learning from Experiments
•  
  9 Evaluation Method and Hypothesis
•  
   9.1 Evaluation of Dialogue Systems
•  
   9.2 Method of Evaluation Cycles
•  
   9.2.1 Experiment Procedure
•  
   9.2.2 Overview of Evaluation Cycles
•  
  10 Evaluation of the Attention-Hesitation Model
•  
   10.1 EC 1: Self-Interruptions as Attention-Regain Strategy
•  
   10.1.1 Attention-Hesitation Dialogue Coordination Model
•  
   10.1.2 Interaction Scenario and Implementation
•  
   10.1.3 Evaluation
•  
   10.1.4 Discussion and Lessons Learned
•  
  10.2 EC 2: Introducing the Focus of Discourse
•  
   10.2.1 Attention-Hesitation Dialogue Coordination Model
•  
   10.2.2 Interaction Scenario and Implementation
•  
   10.2.3 Evaluation
•  
   10.2.4 Discussion and Lessons Learned
•  
  10.3 EC 3: Exploration of a Practical Task during Interaction
•  
   10.3.1 Practical Task: Preparation Phase of Cooking
•  
   10.3.2 Attention-Hesitation Dialogue Coordination Model
•  
   10.3.3 Implementation
•  
   10.3.4 Evaluation
•  
   10.3.5 Discussion and Lessons Learned
•  
  10.4 EC 4: Introducing Lengthening and new Evaluation Approach
•  
   10.4.1 Attention-Hesitation Dialogue Coordination Model
•  
   10.4.2 Interaction Scenario and Implementation
•  
   10.4.3 Evaluation
•  
   10.4.4 Discussion and Lessons Learned
•  
  10.5 EC 5: Bringing It All Together
•  
   10.5.1 Attention-Hesitation Dialogue Coordination Model
•  
   10.5.2 Interaction Scenario and Implementation
•  
   10.5.3 Evaluation
•  
   10.5.4 Discussion and Lessons Learned
•  
  11 Comparison of the Five Evaluation Cycles
•  
   11.1 Features of Attention Concept and Hesitation Strategy
•  
   11.2 Task Performance Hypothesis
•  
   11.2.1 HHI and Attention Hypotheses
•  
   11.2.2 Comparison to other HAI Experiments
•  
   11.3 Side Effects
•  
   11.3.1 Subjective Ratings
•  
   11.3.2 Visual Focus of Attention
•  
   11.4 The Final AHM: Back to the Roots of Attention Theories
•  
   12 Summary of Part III
•  
  13 Conclusion and Perspective
•  
   13.1 Summary of this Thesis
•  
  13.2 Contributions of this Thesis
•  
   13.2.1 Cognitive Interaction Research
•  
   13.2.2 System Engineering Research of Dialogue Modeling
•  
   13.3 Consequences for Smart-Homes
•  
   13.4 Limitations and Future Research Questions
•  
   Bibliography
•  
   A Interfaces of the Dialogue System
•  
  B Questionnaires
•  
   B.1 Task Performance Assessment in EC1
•  
   B.2 Task Performance Assessment in EC2
•  
   B.3 MOS-based Synthesis Evaluation
•  
   B.4 The Godspeed Questionnaire Series
•  
   B.5 Previous Experience Assessment
•  
   B.6 Assessment of Appropriateness of Agent's Statements
•  
  C Study Stimuli
•  
  C.1 Example Stimuli EC1
•  
   C.1.1 Greeting of the Agent
•  
   C.1.2 Information about the Agent
•  
   C.1.3 Farewell of the Agent
•  
   C.2 Example Stimuli EC2
•  
   C.3 Example Stimuli EC4
•  
   C.4 Example Stimuli EC5
•  
   D Instructions for the Participants
•  
   E Data Agreement of the CSRA
•  
   F Additional Analysis EC5
•  
  G Additional Study Results
•  
  G.1 Results of EC1
•  
   G.1.1 Subjective Ratings
•  
   G.1.2 Visual Attention
•  
  G.2 Results of EC2
•  
   G.2.1 Subjective Ratings
•  
   G.2.2 Visual Attention
•  
  G.3 Results of EC3
•  
   G.3.1 Subjective Ratings
•  
  G.4 Results of EC4
•  
   G.4.1 Task Performance
•  
   G.4.2 Visual Attention
•  
   G.4.3 Subjective Ratings
•  
  G.5 Results of EC5
•  
   G.5.1 Subjective Ratings