This book surveys recent advances in Conversational Information Retrieval (CIR), focusing on neural approaches that have been developed in the last few years. Progress in deep learning has brought tremendous improvements in natural language processing (NLP) and conversational AI, leading to a plethora of commercial conversational services that allow naturally spoken and typed interaction, increasing the need for more human-centric interactions in IR.
The book contains nine chapters. Chapter 1 motivates the research of CIR by reviewing the studies on how people search and subsequently defines a CIR system and a reference architecture which is described in detail in the rest of the book. Chapter 2 provides a detailed discussion of techniques for evaluating a CIR system – a goal-oriented conversational AI system with a human in the loop. Then Chapters 3 to 7 describe the algorithms and methods for developing the main CIR modules (or sub-systems). In Chapter 3, conversational document search is discussed, which can be viewed as a sub-system of the CIR system. Chapter 4 is about algorithms and methods for query-focused multi-document summarization. Chapter 5 describes various neural models for conversational machine comprehension, which generate a direct answer to a user query based on retrieved query-relevant documents, while Chapter 6 details neural approaches to conversational question answering over knowledge bases, which is fundamental to the knowledge base search module of a CIR system. Chapter 7 elaborates various techniques and models that aim to equip a CIR system with the capability of proactively leading a human-machine conversation. Chapter 8 reviews a variety of commercial systems for CIR and related tasks. It first presents an overview of research platforms and toolkits which enable scientists and practitioners to build conversational experiences, and continues with historical highlights and recent trends in a range of application areas. Chapter 9 eventually concludes the book with a brief discussion of research trends and areas for future work.
The primary target audience of the book are the IR and NLP research communities. However, audiences with another background, such as machine learning or human-computer interaction, will also find it an accessible introduction to CIR.
Author(s): Jianfeng Gao, Chenyan Xiong, Paul Bennett, Nick Craswell
Series: The Information Retrieval Series, 44
Publisher: Springer
Year: 2023
Language: English
Pages: 216
City: Cham
Preface
Book Organization
Acknowledgments
Contents
1 Introduction
1.1 Related Surveys
1.2 How People Search
1.2.1 Information-Seeking Tasks
1.2.2 Information-Seeking Models
1.3 CIR as Task-Oriented Dialog
1.4 CIR System Architecture
1.4.1 CIR Engine Layer
1.4.2 User Experience Layer
1.4.3 Data Layer
1.4.4 An Example: Macaw
1.5 Remarks on Building an Intelligent CIR System
1.6 Early Works on CIR
1.6.1 System-Oriented and User-Oriented IR Research
1.6.2 System Architecture
1.6.3 Search Result Presentation
1.6.4 Relevance Feedback Interactions
1.6.5 Exploratory Search
2 Evaluating Conversational Information Retrieval
2.1 Forms of Evaluation
2.2 System-Oriented Evaluation
2.2.1 Evaluating Retrieval
2.2.2 Evaluating Retrieval in Conversational Context
2.2.3 Evaluating Non-retrieval Components
2.3 User-Oriented Evaluation
2.3.1 Lab Studies
2.3.2 Scaling Up Evaluation
2.4 Emerging Forms of Evaluation
2.4.1 CIR User Simulation
2.4.2 Responsible CIR
3 Conversational Search
3.1 Task
3.2 Benchmarks
3.2.1 TREC CAsT
3.2.2 OR-QuAC
3.2.3 Other Related Resources
3.3 Pre-trained Language Models
3.3.1 BERT: A Bidirectional Transformer PLM
3.3.2 The Pre-training and Fine-Tuning Framework
3.4 System Architecture
3.4.1 Contextual Query Understanding
3.4.2 Document Retrieval
3.4.3 Document Ranking
3.5 Contextual Query Understanding
3.5.1 Heuristic Query Expansion Methods
3.5.2 Machine Learning-Based Query Expansion Methods
3.5.3 Neural Query Rewriting
3.5.4 Training Data Generation via Rules and Self-supervised Learning
3.6 Sparse Document Retrieval
3.7 Dense Document Retrieval
3.7.1 The Dual-Encoder Architecture
3.7.2 Approximate Nearest Neighbor Search
3.7.3 Model Training
3.8 Conversational Dense Document Retrieval
3.8.1 Few-Shot ConvDR
3.9 Document Ranking
4 Query-Focused Multi-document Summarization
4.1 Task and Datasets
4.2 An Overview of Text Summarization Methods
4.2.1 Extractive Summarizers
Sentence Representation
Sentence Scoring
Summary Sentence Selection
4.2.2 Abstractive Summarizers
FFLMs
RNNs
Transformers
4.3 QMDS Methods
4.3.1 Extractive Methods
Sentence Representation
Sentence Scoring
Summary Sentence Selection
Coarse-to-Fine QMDS
4.3.2 Abstractive Methods
4.4 Factuality Metrics for Summarization Evaluation
5 Conversational Machine Comprehension
5.1 Task and Datasets
5.2 Extractive Readers
5.2.1 BiDAF
5.2.2 BERT-Based Readers
5.3 Generative Readers
5.4 Hybrid Readers
5.5 Conversational Machine Comprehension Readers
5.5.1 History Selection
5.5.2 History Encoding
6 Conversational QA over Knowledge Bases
6.1 Knowledge Bases and Questions
6.1.1 Open Benchmarks
6.2 System Overview
6.3 Semantic Parsing
6.3.1 Overview
6.3.2 A Dynamic Neural Semantic Parser
6.3.3 Using Pre-trained Language Models
6.3.4 C-KBQA Approaches Without Semantic Parsing
6.4 Dialog State Tracking
6.4.1 Contextual Question Rewriting
6.5 Dialog Policy
6.5.1 A Case Study
6.5.2 Dialog Acts
6.5.3 Reinforcement Learning for Policy Optimization
6.6 Response Generation
6.6.1 Template-Based Methods
6.6.2 Corpus-Based Models
6.6.3 SC-GPT
6.7 Grounded Text Generation for C-KBQA
6.7.1 GTG for Task-Oriented Dialog
6.7.2 GTG Training
6.7.3 Remarks on Continual Learning for Conversational Systems
7 Proactive Human-Machine Conversations
7.1 Asking Clarifying Questions
7.1.1 Question Selection Methods
7.1.2 Question Generation Methods
7.2 Suggesting Useful Questions
7.2.1 Question Suggestions in Commercial Search Engines
7.2.2 From Related to Useful Suggestions
7.2.3 Question Suggestion Systems
7.3 Shifting Topics
7.3.1 Topic Shifting in Social Chatbots
7.3.2 Target-Guided Topic Shifting
7.4 Making Recommendations
7.4.1 CRS Architecture
User Experience Layer
Data Layer
CSR Engine Layer
7.4.2 Interactive Preference Elicitation
7.4.3 Recommendation Generation
7.4.4 Explanation Generation
8 Case Study of Commercial Systems
8.1 Research Platforms and Toolkits
8.2 Commercial Applications
8.2.1 Chatbots
Influential Historical Chatbots
Modern Chatbots
8.2.2 Conversational Search Engine
8.2.3 Productivity-Focused Agents
From PAL to Device-Based Assistants
8.2.4 Hybrid-Intelligence Assistants
9 Conclusions and Research Trends
References
