Social Media Processing: 11th Chinese National Conference, SMP 2023, Anhui, China, November 23–26, 2023, Proceedings (Communications in Computer and Information Science)

This book constitutes the thoroughly refereed proceedings of the 11th Chinese National Conference of Social Media Processing, SMP 2023, held in Anhui, China, in November 2023. The 16 full papers presented were carefully reviewed and selected from 88 submissions. The papers are organized in the topical sections on knowledge representation and reasoning; knowledge acquisition and knowledge base construction; linked data, knowledge integration, and knowledge graph storage management; natural language understanding and semantic computing; knowledge graph applications; knowledge graph open resources. Preface Organization Contents DABP: A Domain Augmentation and Bidirectional Stack-Propagation Model for Task-Oriented NLU 1 Introduction 2 Problem Definition 3 The DABP 3.1 Domain Augmented Pre-trained Model 3.2 Bidirectional Stack Propagation Network 4 Experiments 4.1 Experimental Settings 4.2 Baselines 4.3 Overall Results 4.4 Analysis 5 Conclusion References Knowledge Graph Completion via Subgraph Topology Augmentation 1 Introduction 2 Related Works 2.1 Embedding-Based Methods 2.2 Rule-Based Methods 3 Preliminaries 3.1 Definitions 3.2 Problem Formulation 4 Methodology 4.1 Subgraph Extraction and Rule-Based Augmentation 4.2 Subgraph Pruning via Entity Motif Degree 4.3 Completion Scoring and Optimization 5 Experiments 5.1 Experiment Settings 5.2 Link Prediction Results 5.3 Ablation Study 5.4 Parameter Analysis 6 Conclusion References The Diffusion of Vaccine Hesitation: Media Visibility Versus Scientific Authority 1 Introduction 2 Data and Methods 2.1 Datasets 2.2 Methods 3 Results 3.1 Building Datasets Around Individuals 3.2 Comparison of Media Visibility 3.3 Comparison of Scientific Authority 3.4 Scientific Authority and Media Visibility at the Individual Level 3.5 Ways to Engage with Media Visibility 3.6 Media Co-occurrence Networks 3.7 Scientific Inter-citation Networks 3.8 Media Full-Text Text Analysis 4 Conclusion 5 Discussion References An Emotion Aware Dual-Context Model for Suicide Risk Assessment on Social Media 1 Introduction 2 Related Work 3 Methodology 3.1 Post Embedding 3.2 Bi-LSTM Layer 3.3 T-LSTM Layer 3.4 User Features 3.5 Classification Layer 4 Experiment 4.1 Dataset and Evaluation 4.2 Baselines and Hyper-Parameter 4.3 Overall Performance 5 Additional Analyses 5.1 Ablation Study 5.2 SRED Qualitative Analysis 6 Conclusion References Item Recommendation on Shared Accounts Through User Identification 1 Introduction 2 Problem Formulation 3 User Identification 3.1 Heterogeneous Graph Embedding 3.2 Projected Discriminant Attentive Learning 4 Recommendation for Identified User 4.1 SA-BPR: Shared Account-Aware BPR 4.2 Adaptive Preference Adjustment 4.3 Recommendation 5 Experimental Evaluation 5.1 Experimental Setup 5.2 Evaluation of User Identification 5.3 Evaluation of Item Recommendation 6 Conclusion References Prediction and Characterization of Social Media Communication Effects of Emergencies with Multimodal Information 1 Introduction 2 Related Works 2.1 Evaluation of Communication Effects 2.2 Factors Affect Communication Effect 2.3 Multimodal Information in Communication 3 Dataset Construction 4 Communication Effect Prediction Model 4.1 Feature Selection 4.2 Model Architecture 5 Experiment and Evaluation 5.1 Classification Performance 5.2 Feature Importance 6 Conclusion and Future Works References Leverage Heterogeneous Graph Neural Networks for Short-Text Conceptualization 1 Introduction 2 Task Definition 2.1 Concept 2.2 Short-Text Conceptualization 3 Data Modeling 4 Methodology 4.1 Graph Initialization 4.2 Graph Attention Processing 4.3 Optimal Concept Selection 5 Experiments and Results 5.1 Alternative Algorithms 5.2 Datasets 5.3 Experiment Settings 5.4 Experiments on Short-Text Clustering 5.5 Qualitative Analysis 5.6 Experiments on Word Similarity 6 Conclusion References Short-Text Conceptualization Based on Hyper-Graph Learning and Multiple Prior Knowledge 1 Introduction 2 Preliminary 3 Graph Convolutional Network 4 The Proposed Model 4.1 Task Definition 4.2 Hyper-Graph Laplacian Construction 4.3 Hyper-Graph Convolutional Network for Short-Text Conceptualization 4.4 The Variants of the Proposed Model 5 Experiments and Results 5.1 Comparative Models 5.2 Experiments on Concept Quality Evaluation 5.3 Experiments on Ad-Query Similarity Evaluation 6 Conclusion References What You Write Represents Your Personality: A Dual Knowledge Stream Graph Attention Network for Personality Detection 1 Introduction 2 Related Work 3 Methodology 3.1 Task Definition 3.2 Data Preprocessing 3.3 Emotion Dependency Stream 3.4 Psycholinguistic-Emotion Stream 3.5 Classification and Objective 4 Experiments 4.1 Datasets 4.2 Baselines 4.3 Implementation Details 5 Analysis 5.1 Overall Results 5.2 Ablation Study 5.3 Visualization Analysis 5.4 Effect of Trade-Off Parameter 6 Conclusion References Detect Depression from Social Networks with Sentiment Knowledge Sharing 1 Introduction 2 Related Work 2.1 Depression Detection 2.2 Sentiment Analysis 2.3 Multi-Task Learning 3 Methodology 3.1 Input Layer 3.2 Multi-Task Learning Framework 3.3 Gated Attention Layer 3.4 Model Training 4 Experiment 4.1 Datasets 4.2 Baselines and Metrics 4.3 Training Details 4.4 Model Performance 4.5 Ablation Experiments 5 Discussion 6 Conclusion and Future Work References AOM: A New Task for Agitative Opinion Mining in We-media 1 Introduction 2 Related Work 2.1 Opinion Mining 2.2 Offensive Language and Rumor Detection 3 Task Definition 3.1 Agitative Opinion Types 3.2 Problem Formulation 4 Dataset Construction 5 Agitative Opinion Mining 5.1 Baseline Model 5.2 Over-Sampling and Under-Sampling Methods 5.3 Experiment Setup 5.4 Experiment Results 5.5 Error Analysis 6 Conclusion References PNPT: Prototypical Network with Prompt Template for Few-Shot Relation Extraction 1 Introduction 2 Approach 2.1 Prompt Function 2.2 Sentence Encoder 2.3 Prototype Calculation 2.4 Relation Classification 3 Experiments 3.1 Experiment Settings 3.2 Results 3.3 Ablation Study 4 Conclusion References CDBMA: Community Detection in Heterogeneous Networks Based on Multi-attention Mechanism 1 Introduction 2 Related Work 3 Preliminary 4 Proposed Method 4.1 Structure Information Encoder 4.2 Semantic Information Encoder 4.3 Joint Contrastive Optimization 4.4 Experiment 4.5 Conclusion and Further Work References An Adaptive Denoising Recommendation Algorithm for Causal Separation Bias 1 Introduction 2 Related Work 2.1 Denoising 2.2 Debias Recommendation 3 Methodology 3.1 Implicit Feedback Based on Average 3.2 Adaptive Denoising 3.3 Causal Separation Conformity and Interest 4 Experiments 4.1 Experimental Setting and Datasets 4.2 Baselines 4.3 Evaluation 4.4 Data Preprocessing 4.5 Performance of Denoising (RQ1) 4.6 Effect of Causality on Performance (RQ2) 4.7 Regularization Impact (RQ3) 5 Conclusions References Tuning Query Reformulator with Fine-Grained Relevance Feedback 1 Introduction 2 Methodology 2.1 Framework 2.2 Reformulator 2.3 Re-ranking Model 3 Experiments 3.1 Experimental Settings 3.2 Main Results 3.3 Ablation Study 4 Related Work 4.1 Query Reformulation 4.2 Pseudo-Relevance Feedback Methods 4.3 Retrieval-Augmented Model 5 Conclusion References Retrieval-Augmented Document-Level Event Extraction with Cross-Attention Fusion 1 Introduction 2 Related Work 2.1 Document-Level Event Extraction 2.2 Retrieval-Augmented Methods 3 Preliminaries 4 Method 4.1 Retrieval-Augmented Representation 4.2 Candidate Entity Extraction 4.3 Event Detection 4.4 Event Record Generation 4.5 Training and Inference 5 Experiment 5.1 Dataset 5.2 Experiments Setting 5.3 Baselines and Metrics 5.4 Main Results 6 Conclusion References Author Index