Conversational AI for Natural Human-Centric Interaction: 12th International Workshop on Spoken Dialogue System Technology, IWSDS 2021, Singapore

Organization
Preface
Contents
*-1.5pc  Natural Language Understanding
Out-of-Scope Domain and Intent Classification through Hierarchical Joint Modeling
	1 Introduction
	2 Related Work
	3 Hierarchical Joint Modeling
	4 Experiments
		4.1 Experimental Setup
		4.2 Results and Discussion
	5 Conclusion
	References
Segmentation-Based Formulation of Slot Filling Task for Better Generative Modeling
	1 Introduction
	2 Related Work
	3 Segmentation-Based Formulation of Slot Filling Task
		3.1 Definition of Generative Models
		3.2 Training of PYPSMs by Collapsed Gibbs Sampling
		3.3 Finding the Most Likely Labeled Segmentation
	4 Experiment
		4.1 Datasets
		4.2 Settings
		4.3 Results
	5 Conclusion
	References
Can We Predict How Challenging Spoken Language Understanding Corpora Are Across Sources, Languages, and Domains?
	1 Introduction
	2 Predicting Corpus Complexity
	3 Analyzing Complexity Factors
	4 Experiments on Benchmark Corpora
	5 Application to Deployed SLU System Data
	6 Conclusion
	References
*-1.5pc  Personalisation and Generation
Personalized Extractive Summarization with Discourse Structure Constraints Towards Efficient and Coherent Dialog-Based News Delivery
	1 Introduction
	2 Datasets
		2.1 Discourse Structure Dataset
		2.2 Interest Dataset
	3 Methods
		3.1 Inter-Sentence Dependency Parsing
		3.2 Discourse Relation Classification and Chunk Detection
		3.3 Interest Estimation
		3.4 Interesting Document Selection
		3.5 Interesting Sentence Extraction
	4 Experiments
		4.1 Discourse Analysis
		4.2 Interest Estimation
		4.3 Personalized Summarization
	5 Conclusion
	References
Empathetic Dialogue Generation with Pre-trained RoBERTa-GPT2 and External Knowledge
	1 Introduction
	2 Related Work
	3 The Proposed Method
		3.1 Commonsense Knowledge and Emotional Concepts Extractor: CKECE
		3.2 Pre-trained RoBERTa-GPT2 Encoder-Decoder
	4 Experimental Settings and Results Analysis
		4.1 Dataset
		4.2 Baselines
		4.3 Training Details
		4.4 Automatic Evaluation Results
		4.5 Use Cases Study
	5 Conclusion and Outlook
	References
Towards Handling Unconstrained User Preferences in Dialogue
	1 Introduction
	2 Related Work
	3 Data
		3.1 Cambridge Restaurants Dataset
		3.2 Unconstrained Queries
		3.3 Query-Snippets Annotation
	4 Method
		4.1 Unsupervised Approach
		4.2 Supervised Approach
	5 Experiments and Results
		5.1 Information Retrieval
		5.2 Snippet Relevance Classification
		5.3 Discussion
	6 Conclusions
	References
Jurassic is (Almost) All You Need: Few-Shot Meaning-to-Text Generation for Open-Domain Dialogue
	1 Introduction
	2 Datasets
	3 Experimental Setup
	4 Experimental Results
	5 Conclusion
	References
*-1.5pc  Spoken and Multimodal Interaction
Comparison of Automatic Speech Recognition Systems
	1 Introduction
	2 Related Works
	3 EQClinic Dataset
		3.1 Data Collection
		3.2 Data Analysis
	4 Results
	5 Discussion
	6 Conclusion
	References
Multimodal Dialogue Response Timing Estimation Using Dialogue Context Encoder
	1 Introduction
	2 Related Studies
		2.1 Features of Response Timing Estimation
		2.2 Representation of Response Utterance
	3 Multimodal Response Timing Network with Dialogue Context Encoder
		3.1 Inference LSTM
		3.2 Dialogue Context Encoder
	4 Experimental Data
	5 Experiments
		5.1 Feature Extraction
		5.2 Training Condition of Network
	6 Experimental Results
	7 Conclusion
	References
Eliciting Cooperative Persuasive Dialogue by Multimodal Emotional Robot
	1 Introduction
	2 Dialogue Robot with Multimodal Emotional Expressions
		2.1 System Overview
		2.2 Dialogue Scenario
		2.3 Response Selection
		2.4 System's Emotional State
		2.5 System Emotion Decision
		2.6 Speech and Gesture Generation
	3 Speech Corpus for Emotional Dialogue System
		3.1 Response Variation Collection for Each Emotional State
		3.2 Emotional Speech Recording
		3.3 Emotional Robot Gesture
		3.4 Emotion Expressiveness
	4 Dialogue Experiment
		4.1 Experimental Setup
		4.2 Experimental Results on Emotion Effects
		4.3 Experimental Results on Modality Effects
		4.4 Dialogue Example
	5 Conclusion
	References
*-1.5pc  Dialogue Evaluation
Design Guidelines for Developing Systems for Dialogue System Competitions
	1 Introduction
	2 Dialogue System Competitions
		2.1 DSLC2 Situation Track
		2.2 DRC
	3 Proposed Design Guidelines
		3.1 Make the System Take Initiative
		3.2 Prevent Dialogue Flows from Relying Too Much on User Utterances
		3.3 Include in Utterances that the System Understands What the User Said
	4 System Design and Results for DSLC2 Situation Track
		4.1 Specific Designs
		4.2 Results and Examples
	5 System Design for DRC and Results
		5.1 Specific Designs
		5.2 Results of Pre-preliminary Contest and Examples
	6 Discussion and Conclusion
	References
Understanding How People Rate Their Conversations
	1 Introduction
	2 Related Work
	3 Method
		3.1 Personality Story
		3.2 Implementing the Story in a Conversational Agent
	4 Results
		4.1 User Reactions to Personality Story
		4.2 Ratings Through Lens of User Personality
	5 Discussion
	References
*-1.5pc  Dialogue Modelling and Applications
A WoZ Study for an Incremental Proficiency Scoring Interview Agent Eliciting Ratable Samples
	1 Introduction
	2 Related Work
	3 Data Collection
		3.1 Experimental Design
		3.2 WoZ Interview System
		3.3 Interview Data Collection and Human Assessment
	4 Incremental Prediction Model
	5 Results and Discussions
	6 Conclusion
	References
SUPPLE: A Dialogue Management Approach Based on Conversation Patterns
	1 Introduction
	2 Related Work
		2.1 Dialogue Management
	3 The SUPPLE Dialogue Approach
		3.1 Conversational Patterns
		3.2 Session History
		3.3 Agenda and Conversation Management
		3.4 Dialogue State Tracking
		3.5 Dialogue Act Selection
		3.6 Dialogue Systems
	4 Discussion and Conclusion
	References
Dialogue Management as Graph Transformations
	1 Introduction
	2 Related Work
	3 Approach
		3.1 Dialogue State
		3.2 Graph Operations
	4 Case Study
	5 Conclusion
	References
*-1.5pc  Chatbots and Conversational Agent Technologies
Data Collection for Detecting Unwillingness to Answer Questions in Dialogue
	1 Introduction
	2 Dialogue Data Collection
		2.1 Selection of Topics
		2.2 Procedure of Dialogue Data Collection
		2.3 Implementation of Dialogue Data Collection
	3 Collecting Ratings from External Annotators
		3.1 Identifying the Span of Questions and Answers
		3.2 Annotation of Ratings
	4 Estimating Unwillingness to Answer Questions and Difficulty to Ask Questions
		4.1 Procedure for Creating a Model
		4.2 Results of Regression
	5 Summary and Future Work
	References
Enhancing Self-disclosure In Open-Domain Dialogue By Candidate Re-ranking
	1 Introduction
	2 Self-disclosure
		2.1 Self-disclosure (SD) Level
		2.2 Self-disclosure Recognition Model
	3 Self-disclosure Enhancement By Candidate Re-ranking
	4 Experiment
		4.1 Data Configuration
		4.2 Result and Error Analysis
	5 Discussion, Limitation and Future Work
	6 Conclusion
	References
On the Impact of Self-efficacy on Assessment of User Experience in Customer Service Chatbot Conversations
	1 Introduction
	2 Related Work
	3 Experimental Setup
	4 Result
		4.1 Reliability of UX Items
		4.2 UX and Self-efficacy
	5 Discussion and Conclusion
	References
Learning to Ask Specific Questions Naturally in Chat-Oriented Dialogue Systems
	1 Introduction
	2 Related Work
	3 Proposed Model
		3.1 Creating Question-Guiding Corpus
		3.2 Training Response Generation Model
	4 Experiment
		4.1 Question-Guiding Corpus
		4.2 Training Response Generation Model
		4.3 Comparison Models
		4.4 Evaluation by Dialogue Simulation
		4.5 Human Evaluation
	5 Conclusion
	References
Fine-Tuning a Pre-trained Transformer-Based Encoder-Decoder Model with User-Generated Question-Answer Pairs to Realize Character-Like Chatbots
	1 Introduction
	2 Data Collection
		2.1 Role-Play-Based QA Corpus
		2.2 Character-Dialogue Corpus
	3 Response Generation Models
		3.1 Retrieval-Based Model (Baseline)
		3.2 Pre-trained Transformer-Based Encoder-Decoder Model
	4 Experiments
		4.1 Data and Model Preparation
		4.2 Automatic Evaluations
		4.3 Manual Evaluation (Utterance-Level)
		4.4 Manual Evaluation (Dialogue-Level)
	5 Analysis
	6 Conclusion
	References
Investigating the Impact of Pre-trained Language Models on Dialog Evaluation
	1 Introduction
	2 Pre-trained Language Models
		2.1 Masked Language Modeling (MLM)
		2.2 Replaced Token Detection (RTD)
		2.3 Causal Language Modeling (CLM)
		2.4 Permutation Language Modeling (PLM)
		2.5 Sentence-Level Representation Learning
	3 Automatic Dialog Evaluation Metrics
	4 Experiment and Analysis
		4.1 Dialog Evaluation Benchmarks
		4.2 Initial Analysis
		4.3 Rankings of Pre-trained Language Models
		4.4 Fine-Grained Analysis on Evaluation Dimension
	5 Conclusion and Future Work
	References