دانلود کتاب Advances in Information Retrieval: 45th European Conference on Information Retrieval, ECIR 2023, Dublin, Ireland, April 2–6, 2023, Proceedings, Part I

فهرست مطالب :

Preface Organization Keynotes Personalization at Spotify On A Few Responsibilities of (IR) Researchers: Fairness, Awareness, and Sustainability 2022 Karen Spärck Jones Award Lecture Large Language Models for Question Answering: Challenges and Opportunities Contents – Part I Contents – Part II Contents – Part III Full Papers Self-supervised Contrastive BERT Fine-tuning for Fusion-Based Reviewed-Item Retrieval 1 Introduction 2 Background 2.1 IR 2.2 Fusion 2.3 Contrastive Representation Learning 3 Proposed Fusion-based Methods for RIR 3.1 CLFR: Contrastive Learning for Late Fusion RIR 3.2 CEFR: Contrastive Learning for Early Fusion RIR 4 Experiments 4.1 Reviewed-Item Retrieval Dataset (RIRD) 4.2 Experimental Setup 4.3 Baselines 4.4 Evaluation Metrics 4.5 Results and Discussion 5 Conclusion References User Requirement Analysis for a Real-Time NLP-Based Open Information Retrieval Meeting Assistant 1 Introduction 2 Related Work 2.1 Meeting Assistants 2.2 General Guidelines Concerning Recommender Systems 2.3 User Experience of Recommender Systems 3 Method 3.1 User-Centered Design Approach 3.2 Participants 3.3 Study Material 3.4 Study Procedure 4 Results 4.1 Prior Experience with Recommender Systems 4.2 Participants' Experience with the Meeting Assistant 4.3 The Presenters' Experience with the Meeting Assistant 4.4 Envisioned Use Cases for the Meeting Assistant 5 Discussion and Recommendations 5.1 Introduce the System to Users 5.2 Annotate Visual Material with Titles and Keywords 5.3 Cater to User Preferences 5.4 Add Interactivity Options 5.5 Companion Style 6 Limitations and Future Work 7 Conclusion References Auditing Consumer- and Producer-Fairness in Graph Collaborative Filtering 1 Introduction and Motivations 2 Nodes Representation and Neighborhood Exploration in Graph Collaborative Filtering: A Formal Taxonomy 2.1 Preliminaries 2.2 Updating Node Representation Through Message-Passing 2.3 Weighting the Importance of Graph Edges 2.4 Going Beyond Message-Passing 2.5 A Taxonomy of Graph CF Approaches 3 Taxonomy-aware Evaluation 4 Trade-off Analysis 5 Conclusion and Future Work A Experimental Settings and Protocols References Exploiting Graph Structured Cross-Domain Representation for Multi-domain Recommendation 1 Introduction 2 Related Work 3 Methodology 4 Experiments 5 Conclusions and Future Work References Injecting the BM25 Score as Text Improves BERT-Based Re-rankers 1 Introduction 2 Related Work 3 Methods 3.1 First Stage Ranker: BM25 3.2 CECAT: Cross-Encoder Re-rankers Without BM25 Injection 3.3 CEBM25CAT: Cross-Encoder Re-rankers with BM25 Injection 3.4 Linear Interpolation Ensembles of BM25 and CECAT 4 Experimental Design 5 Results 5.1 Main Results: Addressing Our Research Questions 5.2 Analysis of the Results 6 Conclusion and Future Work References Quantifying Valence and Arousal in Text with Multilingual Pre-trained Transformers*-12pt 1 Introduction 2 Related Work 3 Models for Predicting Valence and Arousal from Text 4 Resources 5 Experimental Evaluation 5.1 Results with Different Models and Loss Functions 5.2 Results per Language and Dataset 5.3 Results in Zero-Shot Settings 6 Conclusions and Future Work References A Knowledge Infusion Based Multitasking System for Sarcasm Detection in Meme 1 Introduction 2 Related Work 3 Dataset 4 Methods 4.1 Feature Extraction Layer 4.2 Multimodal Fusion 4.3 Knowledge Infusion (KI) 4.4 Classification 5 Results 6 Analysis 7 Conclusion References Multilingual Detection of Check-Worthy Claims Using World Languages and Adapter Fusion 1 Introduction 2 Related Work 2.1 Identifying Check-Worthy Claims 2.2 Adapters 3 Datasets 3.1 Task Datasets 3.2 Topical Evaluation Dataset 4 Methodology 4.1 World Language Adapter Fusion 4.2 Implementation Details 5 Baselines 6 Results and Discussion 7 Further Analysis 8 Conclusion and Future Work References Market-Aware Models for Efficient Cross-Market Recommendation 1 Introduction 2 Related Work 3 Methodology 3.1 Market-Unaware Models 3.2 Market-Aware Models 4 Experimental Setup 5 Results and Discussion 5.1 Pairwise Experiments 5.2 Global Experiments 6 Conclusions and Future Work References TourismNLG: A Multi-lingual Generative Benchmark for the Tourism Domain 1 Introduction 2 Related Work 2.1 Data Science in Tourism 2.2 Domain-specific Pretrained Models 3 TourismNLG Benchmark 3.1 TourismNLG Datasets 3.2 TourismNLG Tasks 4 Baseline Models for TourismNLG 4.1 Model Selection 4.2 Pre-Training and Finetuning 4.3 Metrics 4.4 Implementation Details for Reproducibility 5 Experiments and Results 6 Conclusions References An Interpretable Knowledge Representation Framework for Natural Language Processing with Cross-Domain Application 1 Introduction 2 Related Work 3 Data Representation Framework 3.1 Tsetlin Machine 3.2 Data Representation 4 Experiments and Results 4.1 Datasets 4.2 Implementation Details 4.3 Baselines 4.4 Results and Analysis 4.5 Visualization 4.6 Concluding Remarks 5 A Case Study: Interpretability 6 Application: Domain Adaptation 7 Conclusion References Graph-Based Recommendation for Sparse and Heterogeneous User Interactions 1 Introduction 2 Related Work 3 Approach 3.1 Heterogeneous Graph Representation of User Interactions 3.2 Generating Recommendations Using Random Walks 3.3 Optimizing Edge Weights Using Genetic Algorithm 4 Experiments 4.1 Use Cases and Datasets 4.2 Evaluation Procedure 4.3 Baselines, Implementation, and Hyperparameters 4.4 Results 5 Conclusions and Future Work References It's Just a Matter of Time: Detecting Depression with Time-Enriched Multimodal Transformers 1 Introduction 2 Related Work 3 Method 4 Experiments 4.1 Datasets 4.2 Experimental Settings 4.3 Comparison Models 4.4 Training and Evaluation Details 5 Results 5.1 Performance Comparison with Prior Works 5.2 Ablation Study 5.3 Error Analysis 5.4 Limitations and Ethical Considerations 6 Conclusion References Recommendation Algorithm Based on Deep Light Graph Convolution Network in Knowledge Graph 1 Introduction 2 Related Work 2.1 Problem Formulation 2.2 Recent Work 3 Method 3.1 Overall Structure of KDL-GCN 3.2 User-Entity Bipartite Graph Method 3.3 Deep Light Graph Convolution Network 3.4 Node Embedding Fusion and Rating Prediction 3.5 Optimization 3.6 Time Complexity Analysis 4 Experiments 4.1 Experiment Setup 4.2 Performance Comparison 4.3 Ablation Analysis 5 Conclusion References Query Performance Prediction for Neural IR: Are We There Yet?*-12pt 1 Introduction 2 Related Work 3 Methodology 4 Experimental Setup 5 Experimental Results 5.1 QPP Models Performance 5.2 ANOVA Analysis 6 Conclusion and Future Work References Item Graph Convolution Collaborative Filtering for Inductive Recommendations 1 Introduction 2 Preliminaries and Related Work 2.1 GCN-Based Recommender 2.2 Item-Based Recommender 3 Methodology 3.1 Graph Projection Module 3.2 Item Embedding Module 3.3 User Embedding Module 3.4 Model Training 4 Experiments 4.1 Datasets 4.2 Baselines 4.3 Transductive Performance 4.4 Inductive Performance 4.5 Ablation Study 5 Conclusion and Future Work References CoLISA: Inner Interaction via Contrastive Learning for Multi-choice Reading Comprehension 1 Introduction 2 Task Formulation 3 Methodology 3.1 DPR-Based Retriever 3.2 In-Sample Attention Mechanism 3.3 Contrastive Learning for Inner Interaction 4 Experimentation 4.1 Experimental Settings 4.2 Main Results 4.3 Analysis 5 Related Work 6 Conclusion References Viewpoint Diversity in Search Results*-12pt 1 Introduction 2 Related Work 3 Evaluating Viewpoint Diversity in Search Results 3.1 Measuring Polarity, Stance, and Logic Bias 3.2 Normalized Discounted Viewpoint Bias 4 Case Study: Evaluating, Fostering Viewpoint Diversity 4.1 Materials 4.2 Viewpoint Diversity Evaluation Results 4.3 Viewpoint Diversification 5 Discussion 6 Conclusion References COILcr: Efficient Semantic Matching in Contextualized Exact Match Retrieval*-12pt 1 Introduction 2 Related Work 3 COILCR: Contextualized Inverted Lists with Canonical Representations 3.1 Term Score Factorization 3.2 Approximate Term Semantic Interaction 4 Experimental Methodology 5 Experiments 5.1 Passage Retrieval Effectiveness 5.2 Balancing Model Efficiency 5.3 Canonical Representation Analysis 6 Conclusion and Future Work References Bootstrapped nDCG Estimation in the Presence of Unjudged Documents*-12pt 1 Introduction 2 Background and Related Work 3 Bootstrapping nDCG Scores 3.1 Preparatory Theoretical Considerations 3.2 Our Bootstrapped nDCG Estimation Approach 3.3 Conceptual Comparison 4 Evaluation 4.1 Experimental Setup 4.2 Evaluation Results 5 Conclusion References Predicting the Listening Contexts of Music Playlists Using Knowledge Graphs*-12pt 1 Introduction 2 Related Work 3 Method 3.1 Matrix Factorisation-Based 3.2 Convolutional Neural Network-Based 3.3 Knowledge Graph-Based 3.4 Hybrid 3.5 Implementation Details 4 Experiments 4.1 Dataset 4.2 Metrics 4.3 Results 5 Conclusions and Future Work References Keyword Embeddings for Query Suggestion*-12pt 1 Introduction 2 Related Work 3 Proposal 4 Experimental Setup 4.1 Datasets 4.2 Baselines 4.3 Evaluation 4.4 Experimental Settings 5 Results 6 Conclusions References Domain-Driven and Discourse-Guided Scientific Summarisation 1 Introduction 2 Related Work 3 Method 3.1 Salient Section Determination 3.2 Rhetorical Content Modelling 3.3 Centrality-based Summariser 4 Experimental Setup 5 Experimental Results 5.1 Structural Discourse Analyses 5.2 Abstract Generation 6 Conclusions References Injecting Temporal-Aware Knowledge in Historical Named Entity Recognition 1 Introduction 2 Related Work 3 Datasets 4 Temporal Knowledge-based Contexts for Named Entity Recognition 4.1 Temporal Information Integration 4.2 Context Retrieval 4.3 Named Entity Recognition Architecture 5 Experimental Setup 5.1 Results 5.2 Impact of Time Intervals 5.3 Impact of Digitization Errors 5.4 Limitations 6 Conclusions & Future Work References A Mask-based Logic Rules Dissemination Method for Sentiment Classifiers 1 Introduction 2 Related Work 2.1 Implicit Methods to Construct Neural-Symbolic Systems 2.2 Explicit Methods to Construct Neural-Symbolic Systems 3 Methodology 3.1 Sources of Logic Rules 3.2 Rule-Mask Mechanism to Disseminate Logical Information 4 Covid-19 Twitter Dataset 4.1 Sentiment Labels 4.2 Rule Labels 4.3 Contrast Labels 4.4 Constructed Dataset 5 Experimental Results 5.1 Dataset Preparation 5.2 Sentiment Classifiers 5.3 Metrics 5.4 Results 6 Conclusion References Contrasting Neural Click Models and Pointwise IPS Rankers 1 Introduction 2 Related Work 3 Background 4 Methods 4.1 Comparing Unbiasedness 4.2 A Difference in Loss Magnitude 5 Experimental Setup 6 Results and Analysis 6.1 Main Findings 6.2 Further Analyses 7 Conclusion References Sentence Retrieval for Open-Ended Dialogue Using Dual Contextual Modeling*-12pt 1 Introduction 2 Related Work 3 Retrieval Framework for Open Dialogues 3.1 Sentence Retrieval Methods 4 Experimental Setting 5 Results 6 Conclusions and Future Work References Temporal Natural Language Inference: Evidence-Based Evaluation of Temporal Text Validity 1 Introduction 2 Related Work 2.1 Temporal Information Retrieval and Processing 2.2 Commonsense Reasoning 2.3 Natural Language Inference 2.4 Incorporation of Knowledge Bases 2.5 Comparison with Related Tasks 3 Task Definition 4 Proposed Method 4.1 Encoding Knowledge 4.2 Combined Model 5 Dataset 5.1 Dataset Construction 5.2 Dataset Statistics 6 Experiments 6.1 Experimental Settings 6.2 Experiments with NLI Pre-training 6.3 Incorporating Common-sense Knowledge 6.4 Testing Different Knowledge Embedding Approaches 7 Conclusion and Future Work References Theoretical Analysis on the Efficiency of Interleaved Comparisons*-12pt 1 Introduction 2 Related Works 2.1 User Click Behavior 2.2 Online Evaluation 3 Preliminary 3.1 Interleaving Method for Analysis (IMA) 3.2 A/B Testing 3.3 Notation 3.4 Definition of Efficiency 4 Theoretical Analysis 5 Numerical Analysis 6 User Simulation 6.1 Datasets 6.2 User Behavior 6.3 Results 7 Conclusion References Intention-Aware Neural Networks for Question Paraphrase Identification 1 Introduction 2 Related Work 3 Preliminary 3.1 VAE, CVAE and MVAE 3.2 Heuristic Intention Extraction 4 Approach 4.1 CVAE-Based Intention-Aware QPI 4.2 MVAE-Based Intention-Aware QPI 5 Experimentation 5.1 Corpora and Evaluation Metrics 5.2 Hyperparameter Settings 5.3 Main Results 5.4 Ablation Experiments 5.5 Effectiveness Analysis 5.6 Case Study 6 Conclusion References Automatic and Analytical Field Weighting for Structured Document Retrieval 1 Introduction 2 Background 3 Information Content Field Weighting (ICFW) 3.1 Model Specification 3.2 Setting the Scale Parameter Lambda 3.3 Approximating Appropriate Values for Lambda Threshold 3.4 Optimising ICFW 4 Experimentation and Analysis 4.1 Data Collections 4.2 Baselines and Methodology 4.3 RQ1: The Effect of Term Frequency Saturation on Performance 4.4 RQ2: Estimating Lambda Analytically 4.5 RQ3: Optimized ICFW Performance 5 Conclusion A Scale Parameter Threshold References An Experimental Study on Pretraining Transformers from Scratch for IR 1 Introduction 2 Related Work 3 Pretraining from Scratch 3.1 Pretraining 4 Experiments 4.1 RQ1: Are Models Fully Trained on MSMARCO as Good as Models Pretrained on a Diverse Collection Set? 4.2 RQ2: Do Models Pretrained in MSMARCO Generalize Well on Other Collections? 4.3 RQ3: Can We Take Advantage of that Pretraining from Scratch in Collections of Specialized Domains/Languages 4.4 RQ4: Impact of Architectures 5 Conclusion References Neural Approaches to Multilingual Information Retrieval*-12pt 1 Introduction 2 Background 3 Fine-Tuning MPLMs for MLIR 3.1 English Training (ET) 3.2 Multilingual Translate Training (MTT) 4 Experiments 4.1 Neural Retrieval Models 4.2 Evaluation 5 Results 5.1 Multilingual Batching for Fine-Tuning 5.2 Effectiveness Relative to Baselines 5.3 Preprocessing and Indexing Time 6 Analysis 6.1 Language Bias 6.2 Example Queries 7 Conclusion and Future Work A MTT Implementation Details References CoSPLADE: Contextualizing SPLADE for Conversational Information Retrieval*-12pt 1 Introduction 2 Related Works 3 Model 3.1 First Stage 3.2 Reranking 4 Experimental Protocol 4.1 Datasets 4.2 Metrics and Baselines 5 Results 5.1 First-Stage Ranking Effectiveness 5.2 Second-Stage Ranking Effectiveness 5.3 Effectiveness Compared to TREC CAsT Participants 5.4 Efficiency 6 Conclusion References SR-CoMbEr: Heterogeneous Network Embedding Using Community Multi-view Enhanced Graph Convolutional Network for Automating Systematic Reviews 1 Introduction 2 Preliminaries 2.1 Heterogeneous Information Network 2.2 Graph Convolutional Networks 3 Related Works 4 SR-CoMbEr 4.1 Heterogeneous Community Detection 4.2 Community Multi-view Learning 4.3 Global Consensus 5 Experimental Design 5.1 Dataset 5.2 Baselines 5.3 Evaluation Metrics 5.4 Implementation Details 6 Experimental Results 6.1 Systematic Review 6.2 Ablation Study 7 Conclusion References Multimodal Inverse Cloze Task for Knowledge-Based Visual Question Answering 1 Introduction 2 Related Work 3 Methods 3.1 Information Retrieval Framework 3.2 Models 3.3 Training Stages 3.4 Inference 3.5 Implementation Details 4 Results 4.1 Information Retrieval 4.2 Reading Comprehension 5 Generic vs. Specialized Image Representations 6 Conclusion and Perspectives References A Transformer-Based Framework for POI-Level Social Post Geolocation 1 Introduction 2 Related Work 2.1 Post Geolocation 2.2 Hierarchical Geolocation 3 Method 3.1 Problem Formulation 3.2 Method Overview 3.3 Feature Representation 3.4 Feature Fusion 3.5 Hierarchical Prediction 4 Experimental Setting 4.1 Datasets 4.2 Evaluation Metrics 4.3 Parameter Setting 4.4 Baselines 5 Experimental Results 5.1 Baseline Comparison 5.2 Representation Combination Selection 5.3 Ablation Study 5.4 Coarse-Level Geolocation 6 Conclusion References Document-Level Relation Extraction with Distance-Dependent Bias Network and Neighbors Enhanced Loss 1 Introduction 2 Methodology 2.1 Crossing-Distance Calculation 2.2 SDBN 2.3 Bias Network 2.4 Neighbors Enhanced Loss 3 Experiment 3.1 Datasets 3.2 Implementation Details 3.3 Compared Methods 3.4 Main Results 3.5 Ablation Study 3.6 Robustness Analysis 3.7 Hyper-parameter Analysis 3.8 Case Study 4 Related Work 5 Conclusion References Investigating Conversational Agent Action in Legal Case Retrieval 1 Introduction 2 Related Work 3 User Study 3.1 Conversational Legal Case Retrieval 3.2 Tasks and Participants 3.3 Procedure 4 Results 4.1 Analysis on Conversational Agent Action 4.2 Conversational Agent Action Prediction 5 Conclusion References MS-Shift: An Analysis of MS MARCO Distribution Shifts on Neural Retrieval 1 Introduction 2 Related Work 3 Methodology 3.1 Distribution Shifts 3.2 Evaluation Procedure 4 Experimental Setup 5 Results and Analysis 5.1 Performance Evaluation on Distribution Shifts 5.2 Train/Test Distribution Similarity 6 Conclusion References Listwise Explanations for Ranking Models Using Multiple Explainers*-12pt 1 Introduction 2 Related Work 3 Background and Preliminaries 3.1 Explainers for Ranking 3.2 Explanations to a Ranking Model 3.3 Problem Statement 4 Generalized Preference Coverage 4.1 The Preference Coverage Framework 4.2 Optimizing PC for Multiple Explainers 5 Experimental Setup 5.1 Datasets and Ranking Models 5.2 Baseline and Competitors 5.3 Metrics 6 Evaluation Results 6.1 Effectiveness of Explanations 6.2 Utility of Explanations 7 Conclusion and Outlook References Improving Video Retrieval Using Multilingual Knowledge Transfer 1 Introduction 2 Related Work 2.1 Video Retrieval 2.2 Multilingual Training 3 MKTVR: Multilingual Knowledge Transfer for Video Retrieval 3.1 Problem Statement 3.2 Approach 4 Experiments 4.1 Datasets 4.2 Metrics 4.3 Implementation Details 5 Results and Discussion 5.1 Evaluation on English Video Retrieval Datasets 5.2 Evaluation on Multilingual Video Retrieval Datasets 5.3 Ablation Studies 6 Conclusion References Service Is Good, Very Good or Excellent? Towards Aspect Based Sentiment Intensity Analysis 1 Introduction 2 Resource Creation 3 Methodology 3.1 Seq2Seq Generation 3.2 Model Explainability 4 Experiments and Analysis 4.1 Experimental Results 4.2 Detailed Analysis 5 Conclusion References Effective Hierarchical Information Threading Using Network Community Detection*-12pt 1 Introduction 2 Related Work 3 Proposed Approach: HINT 4 Experimental Setup 5 Offline Evaluation 5.1 Results 5.2 Ablation Study 6 User Study 6.1 Results 7 Identifying Incremental Threads 8 Conclusions References HADA: A Graph-Based Amalgamation Framework in Image-Text Retrieval 1 Introduction 2 Related Work 3 Methodology 3.1 Revisit State-of-the-Art Models 3.2 Create Graph Structure 3.3 Graph Neural Network 3.4 Training Tasks 4 Experiment 4.1 Dataset and Evaluation Metrics 4.2 Implementation Details 4.3 Baselines 4.4 Comparison to Baseline 4.5 HADA with Other Input Models 5 Conclusion References Author Index