Web and Big Data: 5th International Joint Conference, APWeb-WAIM 2021, Guangzhou, China, August 23–25, 2021, Proceedings, Part I (Lecture Notes in Computer Science)

Description

This two-volume set, LNCS 12858 and 12859, constitutes the thoroughly refereed proceedings of the 5th International Joint Conference, APWeb-WAIM 2021, held in Guangzhou, China, in August 2021. The 44 full papers presented together with 24 short papers, and 6 demonstration papers were carefully reviewed and selected from 184 submissions. The papers are organized around the following topics: Graph Mining; Data Mining; Data Management; Topic Model and Language Model Learning; Text Analysis; Text Classification; Machine Learning; Knowledge Graph; Emerging Data Processing Techniques; Information Extraction and Retrieval; Recommender System; Spatial and Spatio-Temporal Databases; and Demo. Preface Organization Keynotes Approaches to Distributed RDF Data Management and SPARQL Processing Striving for Socially Responsible AI in Data Science Democratizing the Full Data Analytics Software Stack Efficient Network Embeddings for Large Graphs Contents – Part I Contents – Part II Graph Mining Co-authorship Prediction Based on Temporal Graph Attention 1 Introduction 2 Related Work 3 Problem Description 4 Our Approach 4.1 Semantic Encoder 4.2 Embedding Encoder 4.3 Decoder 4.4 Optimization 5 Experiments 5.1 Datasets 5.2 Training Protocol 5.3 Evaluation Protocol 5.4 Results and Analysis 5.5 Ablation Study 6 Conclusion and Future Work References Degree-Specific Topology Learning for Graph Convolutional Network 1 Introduction 2 Related Work 3 Notations and Preliminaries 3.1 Notations 3.2 Graph Convolutional Network 4 Degree-Specific Topology Learning for GCNs 4.1 Methodology 4.2 Remove Edges for High-Degree Nodes 4.3 Add Edges for Low-Degree Nodes 4.4 Framework 5 Experiments 5.1 Datasets and Baselines 5.2 Comparison with the Existing Methods 5.3 Ablation Experiments 5.4 Effect of Nodes with Different Degree 5.5 Parameters Sensitiveness 5.6 Classification Performance with Deeper GCN 6 Conclusion References Simplifying Graph Convolutional Networks as Matrix Factorization 1 Introduction 2 GCN as Unitized and Co-training MF 3 The UCMF Architecture 4 Experiments 4.1 Experimental Settings 4.2 Results and Analysis 5 Conclusion References GRASP: Graph Alignment Through Spectral Signatures 1 Introduction 2 Related Work 3 Background and Problem 4 GRASP 4.1 Choice of Basis: Normalized Laplacian 4.2 Choice of Functions: Heat Kernel 4.3 Mapping Matrix 4.4 Node-to-Node Correspondence 4.5 Base Alignment 5 Experiments 6 Conclusion References FANE: A Fusion-Based Attributed Network Embedding Framework 1 Introduction 2 Related Works 3 Proposed Method 3.1 Problem Definition 3.2 Overall Framework 4 Experiments 4.1 Experimental Settings 4.2 Experiment Results 5 Conclusion and Further Work References Data Mining What Have We Learned from OpenReview? 1 Introduction 2 Dataset 3 Results Learned from Open Reviews 3.1 How Is the Impact of Non-expert Reviewers? 3.2 Which Aspects Play Important Roles in Review Score? 3.3 Which Research Field Has Higher/Lower Acceptance Rate? 3.4 Review Score Vs. Citation Number 3.5 Do Submissions Posted on ArXiv Have Higher Acceptance Rate? 4 Related Work 5 Conclusion References Unsafe Driving Behavior Prediction for Electric Vehicles 1 Introduction 2 Related Work 3 Preliminary 3.1 Electric Vehicle Dataset 3.2 Problem Formulation 3.3 Characteristics of Driving Behaviors 4 Methods 4.1 Overview of the Proposed Approach 4.2 Driving Behavior Feature Representation 4.3 Behavior Prediction Method 4.4 Types of Unsafe Driving Behaviors 5 Experimental Study 5.1 Experimental Settings 5.2 Evaluation of Classification Performance 5.3 Behavior Type Prediction 5.4 Feature Importance Evaluation 6 Conclusions References Resource Trading with Hierarchical Game for Computing-Power Network Market 1 Introduction 2 The Proposed Framework Description 2.1 Users Layer 2.2 Computing-Power Provider Layer 2.3 Computing Network Service Provider Layer 3 System Model and Problem Formulation 3.1 System Model 3.2 Problem Formulation 3.3 Game Analysis 4 Multi-agent Reinforcement Learning Algorithm 4.1 Multi-agent Model 4.2 Multi-agent Reinforcement Learning 5 Simulation Results and Discussions 5.1 Experimental Setting and Algorithm Convergence 5.2 Performance of DG-RL Algorithm 5.3 Impacts of the Number of Users 5.4 Impacts of the Congestion Coefficient 6 Conclusions and Future Work References Analyze and Evaluate Database-Backed Web Applications with WTool 1 Introduction 2 Background and Motivation 2.1 Design of Spring Applications 2.2 Challenges Faced by Developers 3 Design 3.1 SQL Collector 3.2 Semantic Analyzer 3.3 Benchmark Script Generator 4 Analyze the Web Applications 4.1 Single Query 4.2 Transaction 4.3 Database Tables 5 Evaluation with WTool 5.1 Optimize Pagination 5.2 Eliminate Redundant Explicit Transaction 6 Related Work 7 Limitation 8 Conclusion References Semi-supervised Variational Multi-view Anomaly Detection 1 Introduction 2 Preliminaries 3 Methodology 3.1 Problem Setting and Proposed Framework 3.2 Loss Function Deduction 3.3 Categorical Distribution for Discrete Data 3.4 Semi-supervised Multi-view Anomaly Detection Score Design 4 Experiments 5 Conclusion References A Graph Attention Network Model for GMV Forecast on Online Shopping Festival 1 Introduction 2 The Proposed Model 2.1 Graph Neural Network Based Encoder 2.2 Two-Way Regression Decoder 3 Experiment 3.1 Experimental Setup 3.2 Result Analysis 4 Conclusion References Suicide Ideation Detection on Social Media During COVID-19 via Adversarial and Multi-task Learning 1 Introduction 2 Related Work 3 Methodology 4 Experiments 4.1 Dataset 4.2 Performance on the SID Dataset 4.3 Ablation Study 4.4 Importance of Emotion Feature 5 Conclusion References Data Management An Efficient Bucket Logging for Persistent Memory 1 Introduction 2 Background 2.1 Transactions for Database Systems 2.2 Log-Based Recovery Algorithms 3 Experimental Study on Optane DCPMM 4 Bucket Logging 4.1 System Overview 4.2 Data Structure of Log Bucket 4.3 Commit Protocol 4.4 Recovery Protocol 5 Evaluation 5.1 Experimental Platform 5.2 Sensitivity Analysis 5.3 Runtime Performance 5.4 Recovery Performance 6 Related Work 7 Conclusion and Future Work References Data Poisoning Attacks on Crowdsourcing Learning 1 Introduction 2 Problem Setting 3 Data Poisoning Attack Method 3.1 Our Adversarial Strategy 3.2 Computing Our Attack Strategy 4 Experiments 4.1 Datasets and Baselines 4.2 Results and Analysis 5 Related Work 6 Conclusions References Dynamic Environment Simulation for Database Performance Evaluation 1 Introduction 2 Related Work 3 Workload Generator Definition 3.1 Workload Generator Definition 4 Environment Simulation 4.1 Workload Modeling on Each Individual Dimension 4.2 Modeling Environment by Learning Workload Interaction Among Dimensions 4.3 Dynamic Environment Simulation 5 Experiment Results 5.1 Environment Workload Demonstration 5.2 Environment Simulation 5.3 Environment Simulation Based on Real Applications 6 Conclusion References LinKV: An RDMA-Enabled KVS for High Performance and Strict Consistency Under Skew 1 Introduction 2 Background and Related Work 2.1 Skew Mitigation Under ccNUMA Abstraction 2.2 Enforcing Strict Consistency 3 LinKV Design 4 Write 4.1 Sequential Buffer Coordination 5 Read 5.1 Read Consistency Challenge 5.2 Lease Strategy 5.3 Lease-Based Read 6 Experiment 7 Conclusion References Cheetah: An Adaptive User-Space Cache for Non-volatile Main Memory File Systems 1 Introduction 2 Background 2.1 Non-volatile Main Memory 2.2 Direct Access and Memory Mapping 3 Design 3.1 CUlib 3.2 Hybrid Memory Management 4 Cache Replacement 5 Evaluation 5.1 Experimental Setup 5.2 Microbenchmarks 5.3 Redis 6 Related Work 7 Conclusion References Topic Model and Language Model Learning Chinese Word Embedding Learning with Limited Data 1 Introduction 2 Related Work 3 Methodology 3.1 Optimization Objective 3.2 Parameter Inference 3.3 Time and Space Complexity 4 Experiments 4.1 Datasets 4.2 Baselines and Settings 4.3 Result Analysis 5 Conclusion References Sparse Biterm Topic Model for Short Texts 1 Introduction 2 Related Work 2.1 Topic Models over Short Texts 2.2 Sparse Topic Model 3 Our Method 3.1 A Brief Review of BTM 3.2 SparseBTM 4 Experiments 4.1 Datasets 4.2 Evaluation Methods 4.3 Baseline and Parameter Settings 4.4 Experimental Results 5 Conclusion References EMBERT: A Pre-trained Language Model for Chinese Medical Text Mining 1 Introduction 2 Related Work 2.1 Pre-trained Language Models in the Open Domain 2.2 Pre-trained Language Models in Medical Domain 3 The EMBERT Model 3.1 Context-Entity Consistency Prediction 3.2 Entity Segmentation 3.3 Bidirectional Entity Masking 3.4 Overll Loss Function 4 Experiments 4.1 Experimental Settings 4.2 Baseline Models and Downstream Task Datasets 4.3 Overall Model Results 4.4 Ablation Studies 4.5 Analysis of Attention Weight Distributions 4.6 Varying the Corruption Rate 5 Conclusion and Future Work References Self-supervised Learning for Semantic Sentence Matching with Dense Transformer Inference Network 1 Introduction 2 Related Work 2.1 Self-supervised Learning 2.2 Semantic Sentence Matching 3 Our Approach 3.1 Task Definition 3.2 Encoding with Self-supervised Learning 3.3 Inference-Attention Block 3.4 Densely-Connected Inference Block 3.5 Global Average Logits 4 Experiments and Analysis 4.1 Implementation Details 4.2 Effectiveness of Densely Connect and GAL 4.3 Effectiveness of Inference-Attention 4.4 Effectiveness of Self-supervised Learning 4.5 Experiments on Paraphrase Identification 4.6 Experiments on Natural Language Inference 5 Conclusion References An Explainable Evaluation of Unsupervised Transfer Learning for Parallel Sentences Mining 1 Introduction 2 Evaluation for Transfer Learning 3 Unsupervised Transfer Learning for Mining Parallel Sentences 3.1 Language Selector 3.2 Transfer Learning for Mining Parallel Data 4 Experimental Setting 5 Results and Discussions 5.1 Empirical Evaluation of GH Distance 5.2 Results on BUCC 5.3 Results on Low-Resource Language Pair 6 Conclusion References Text Analysis Leveraging Syntactic Dependency and Lexical Similarity for Neural Relation Extraction 1 Introduction 2 Preliminary 2.1 Problem Definition 2.2 Definition 1: Concept 2.3 Definition 1: Instance Conceptualization 3 Methodology 3.1 Input Generation 3.2 Sentence-Level Semantic Vector Generation 3.3 Entity-Level Semantic Vector Generation 3.4 Lexical Semantic Similarity Based on Concept 3.5 DS-ATT: Attention Based on Syntactic Dependency and Lexical Similarity 3.6 Objective 4 Experiments 4.1 Datasets and Metric 4.2 Comparative Models 4.3 Experimental Results 5 Conclusion References A Novel Capsule Aggregation Framework for Natural Language Inference 1 Introduction 2 Related Work 3 Proposed Method 3.1 Word Representation 3.2 Matching Block 3.3 Capsule Aggregation Block 4 Experiments and Analyses 4.1 Model Training 4.2 Overall Performance 4.3 Ablation Experiments 4.4 Case Study 5 Conclusion References Learning Modality-Invariant Features by Cross-Modality Adversarial Network for Visual Question Answering 1 Introduction 2 Related Work 3 Methodology 3.1 Features Representation 3.2 Information Integration 3.3 Cross-Modality Adversarial Learning 3.4 Modality-Invariant Attention Learning 3.5 Feature Fusion and Answer Prediction 3.6 Optimization 4 Experiments 4.1 Datasets 4.2 Implementation Details 4.3 Result Analysis 5 Conclusion References Difficulty-Controllable Visual Question Generation 1 Introduction 2 Related Work 2.1 Text Question Generation 2.2 Visual Question Generation 3 Our Proposed Framework 3.1 Problem Definition 3.2 Image Encoder 3.3 Answer Encoder 3.4 Fusion Module 3.5 Difficulty-Controllable Decoder 3.6 Training and Inference 4 Experimental Settings 4.1 Training Data Construction 4.2 Experimental Details 4.3 Evaluation Metrics 4.4 Baselines and Ablation Tests 4.5 Results of Visual Question Generation 4.6 Difficulty Control Results 4.7 Human Evaluation 4.8 Case Study 5 Conclusion References Incorporating Typological Features into Language Selection for Multilingual Neural Machine Translation 1 Introduction 2 Methodology 2.1 Overall Architecture 2.2 Features 2.3 Language Selection Model 3 Experimental Setup 3.1 Data 3.2 Settings 3.3 Results and Analysis 4 Conclusion References Removing Input Confounder for Translation Quality Estimation via a Causal Motivated Method 1 Introduction 2 Approach 2.1 Eliminating Confounder Information via Half-Sibling Regression 2.2 Finding and Eliminating Confounder Information 3 Experiments 3.1 Main Results 3.2 Distribution of the Eligible Features Under the Denoising Condition 4 Conclusion and Future Work References Text Classification Learning Refined Features for Open-World Text Classification 1 Introduction 2 Related Work 3 Model 3.1 Problem Definition 3.2 Model Framework 3.3 Feature Regularization via Class Descriptions 4 Experiments 4.1 Compared Methods 4.2 Datasets 4.3 Class Descriptions 4.4 Experimental Settings 4.5 Results and Analysis 5 Discussion and Conclusion References Emotion Classification of Text Based on BERT and Broad Learning System 1 Introduction 2 Related Work 3 Modeling on Emotion Classification Based on BERT and BLS 3.1 Overview of Model Structure 3.2 BERT Pre-trained Model for Sentence Embedding 3.3 BLS for Emotion Classification of Text 3.4 Incremental Learning of BLS 3.5 Three Cascading Structures of BLS 4 Experiment 4.1 Dataset 4.2 Baseline Methods 4.3 Parameter Setting 4.4 Results and Analysis 5 Conclusion and Future Research References Improving Document-Level Sentiment Classification with User-Product Gated Network 1 Introduction 2 Related Work 3 Methodology 3.1 The UP-LSTM Cell 3.2 Overall Architecture 3.3 Coupling Input and Forget Gates 4 Evaluation 4.1 Experiment Setup 4.2 Results and Analysis 4.3 Visualization 5 Conclusion References Integrating RoBERTa Fine-Tuning and User Writing Styles for Authorship Attribution of Short Texts 1 Introduction 2 Related Work 3 Methodology 3.1 Text Representation Module 3.2 User Writing Style Module 3.3 Combination of Two Modules 4 Experiments 4.1 Experiment Setup 4.2 Authorship Attribution Results 5 Conclusion and Future Work References Dependency Graph Convolution and POS Tagging Transferring for Aspect-Based Sentiment Classification 1 Introduction 2 Methodology 3 Experiments 3.1 Datasets and Experimental Settings 3.2 Experimental Analysis 4 Conclusions and Future Work References Machine Learning 1 DTWSSE: Data Augmentation with a Siamese Encoder for Time Series 1 Introduction 2 Related Work 3 The Proposed DTWSSE Method 3.1 Make the Dataset Balanced 3.2 Use DTW to Select Instances for New Data Generation 3.3 Interpolation Adapted to the DTW Metric 4 Experiment 4.1 Basic Results 4.2 Ablation Studies 4.3 Apply to Balanced Datasets 5 Conclusion References PT-LSTM: Extending LSTM for Efficient Processing Time Attributes in Time Series Prediction 1 Introduction 2 Related Work 3 Method 3.1 Problem Description 3.2 Position Encoding of Time Attributes 3.3 PT-LSTM Model 3.4 Time Series Prediction Based on PT-LSTM 4 Experiments 4.1 Datasets and Metrics 4.2 Experimental Settings 4.3 Experimental Results 5 Conclusion References Loss Attenuation for Time Series Prediction Respecting Categories of Values 1 Introduction 2 Related Work 2.1 Loss Functions and Their Weights 2.2 Time Series Prediction Based on Neural Networks 3 Time Series Forecasting with Categories 4 Category-Aware Adaptive Loss Attenuation 4.1 Boundary Closeness of Samples to Category Effective Areas 4.2 The Prediction Framework 5 Experimental Evaluation 5.1 Data Set 5.2 Experiment Settings 5.3 Experimental Results 6 Conclusion References PFL-MoE: Personalized Federated Learning Based on Mixture of Experts 1 Introduction 2 Methodology 2.1 PFL-MoE 2.2 PFL-MF and PFL-MFE 3 Experimental Evaluation 3.1 Experimental Setup 3.2 Personalization Effect of PFL-MF and PFL-MFE 4 Conclusion and Future Work References A New Density Clustering Method Using Mutual Nearest Neighbor 1 Introduction 2 The RDC Algorithm 2.1 Related Work: Natural Neighbor Structure 2.2 Relative Density 2.3 The RDC Clustering Algorithm 2.4 Complexity Analysis 3 Experiments and Results Analysis 4 Conclusions References Author Index