Knowledge Science, Engineering and Management: 16th International Conference, KSEM 2023, Guangzhou, China, August 16–18, 2023, Proceedings, Part IV (Lecture Notes in Artificial Intelligence)

This volume set constitutes the refereed proceedings of the 16th International Conference on Knowledge Science, Engineering and Management, KSEM 2023, which was held in Guangzhou, China, during August 16–18, 2023. The 114 full papers and 30 short papers included in this book were carefully reviewed and selected from 395 submissions. They were organized in topical sections as follows: knowledge science with learning and AI; knowledge engineering research and applications; knowledge management systems; and emerging technologies for knowledge science, engineering and management. Preface Organization Keynotes Abstracts Credibility of Machine Learning Through Information Granularity A New Paradigm to Leverage Formalized Knowledge and Machine Learning Recent Advances in Assessing Time Series Similarity Through Dynamic Time Warping ChatGLM: Run Your Own “ChatGPT” on a Laptop Contents – Part IV Emerging Technologies for Knowledge Science, Engineering and Management Federated Prompting and Chain-of-Thought Reasoning for Improving LLMs Answering 1 Introduction 2 Related Work 3 Scenarios and Approaches 3.1 Synonymous Questions with Same Parameters (SP-Questions) 3.2 Synonymous Questions with Different Parameters (DP-Questions) 4 Experiment 4.1 Results of Fed-SP-SC 4.2 Results of Fed-DP-CoT 4.3 Ablation Studies 5 Conclusion References Advancing Domain Adaptation of BERT by Learning Domain Term Semantics 1 Introduction 2 Related Work 3 Methodology 3.1 Semantic Acquisition 3.2 Semantic Injection 4 Experiments 4.1 Experimental Setup 4.2 Baselines 4.3 Experimental Results 4.4 Ablation 5 Conclusion References Deep Reinforcement Learning for Group-Aware Robot Navigation in Crowds 1 Introduction 2 Related Work 2.1 Robots Navigating Through Crowds 2.2 Group Recognition in Crowded Environments 3 Method 3.1 Problem Formulation 3.2 Group Space Recognition 3.3 DRL Framework Combining Group Features 3.4 Network Architecture 4 Experiments 4.1 Experimental Setup 4.2 Results and Analysis 5 Summary References An Enhanced Distributed Algorithm for Area Skyline Computation Based on Apache Spark 1 Introduction 2 Related Work 2.1 Skyline Query 2.2 Skyline Computation with Distributed Framework 2.3 Area Skyline Computation with MapReduce 3 Area Skyline Computation with Spark 3.1 Local Partial Skyline Extraction 3.2 Filter Creation 3.3 Filtering 4 Experiment 4.1 Baselines 4.2 Effect of Grids and Facilities 5 Conclusion References TCMCoRep: Traditional Chinese Medicine Data Mining with Contrastive Graph Representation Learning 1 Introduction 2 Related Work 2.1 Graph Convolution and Graph Neural Network 2.2 Contrastive Learning 3 Preliminary 4 Framework 4.1 Heterogeneous Encoding 4.2 Graph Augmentation 4.3 Homogeneous Encoding 4.4 Momentum Mechanism 4.5 GinfoNCE Loss 5 Experiments 5.1 Datasets 5.2 Syndrome Detection 5.3 Prescription Generation 5.4 Node Classification 5.5 Ablation Study 5.6 Case Study 6 Conclusion References Local-Global Fusion Augmented Graph Contrastive Learning Based on Generative Models 1 Introduction 2 Preliminaries 3 Method 3.1 Overview 3.2 Graph Diffusion 3.3 Pre-training Generative Models 3.4 Encoders 3.5 Training 4 Experiments 4.1 Experimental Setup 4.2 Node Classification 4.3 Node Clustering 4.4 Visualization of Embeddings 4.5 Ablation Study 5 Related Work 5.1 Contrastive Representation Learning 5.2 Graph Diffusion Network 6 Conclusions References PRACM: Predictive Rewards for Actor-Critic with Mixing Function in Multi-Agent Reinforcement Learning 1 Introduction 2 Background 2.1 Dec-POMDP 2.2 QMIX and QPLEX 2.3 MADDPG and FACMAC 2.4 MACOPT and ELIGN 3 Method 3.1 Training Critic 3.2 Policy Gradients 3.3 Predictive Rewards 4 Experiments and Results 4.1 Cooperative Predator-Prey 4.2 SMAC 5 Conclusion References A Cybersecurity Knowledge Graph Completion Method for Scalable Scenarios 1 Introduction 2 Related Work 2.1 Research on Knowledge Graph Completion 2.2 Research on Meta-Learning 3 Methodology 3.1 Meta-Knowledge Learning 3.2 Scoring Function 3.3 Robust Representation Learning on Samples 3.4 SEMA Self-distillation 4 Experiments 4.1 Dataset 4.2 Evaluation Metrics and Results 5 Conclusion References Research on Remote Sensing Image Classification Based on Transfer Learning and Data Augmentation 1 Introduction 2 The Resnet50 Model Based on Transfer Learning 2.1 Analysis of the ResNet50 Model 2.2 Log Softmax Classifier 2.3 Classification Methods of Transfer Learning Models 2.4 Data Augmentation 2.5 ResNet50-TL 2.6 Evaluation Metrics 3 Model Training and Validation 3.1 Experimental Dataset 3.2 Data Pre-processing 3.3 Image Data Augmentation and Feature Extraction Figure 3.4 Training and Testing of Remote Sensing Picture Classifier Models 4 Conclusion References Multivariate Long-Term Traffic Forecasting with Graph Convolutional Network and Historical Attention Mechanism 1 Introduction 2 Related Work 2.1 Traditional Time Series Forecasting 2.2 Multivariate Spatial-Temporal Short-Term Series Forecasting 2.3 Multivariate Long-Term Time Series Forecasting 3 Methodology 3.1 Problem Definition 3.2 Historical Attention Mechanism 3.3 Temporal Graph Convolutional Network 3.4 Spatial Graph Convolutional Network 3.5 Framework 4 Experiments 4.1 Baselines 4.2 Datasets 4.3 Experimental Setups 4.4 Main Results 4.5 Ablation Studies 5 Conclusion References Multi-hop Reading Comprehension Learning Method Based on Answer Contrastive Learning 1 Introduction 2 Related Work 2.1 Multi-hop RC 2.2 Contrastive Learning 3 Model 3.1 Data Augmentation 3.2 Shared Encoder 3.3 Answer Contrastive Learning 3.4 Graph Neural Network 3.5 Multi-task Joint Learning 4 Experiments and Analyses 4.1 Dataset 4.2 Baselines and Metrics 4.3 Experiment Settings 4.4 Experimental Results 4.5 Ablation Study 4.6 Hyperparameter Analysis 4.7 Case Study 5 Conclusion References Importance-Based Neuron Selective Distillation for Interference Mitigation in Multilingual Neural Machine Translation 1 Introduction 2 Background 3 Method 3.1 Pruning 3.2 Selective Knowledge Distillation 3.3 Fine-Tuning 4 Experiments 4.1 Data Preparation 4.2 Model Details and Baselines 4.3 Main Results 4.4 Ablation Study 4.5 Analysis 5 Conclusion 6 Limitations and Future Work References Are GPT Embeddings Useful for Ads and Recommendation? 1 Introduction 2 Related Works 2.1 Text Semantic Modeling 2.2 Large Language Models 3 Methodology 3.1 Overall Framework 3.2 Embedding as a Feature (EaaF) 3.3 Embedding as a Regularization (EaaR) 3.4 Embedding as a Pre-traning Task (EaaP) 4 Experiments 4.1 Datasets and Baselines 4.2 Methods Performance 4.3 Impact of Backbone 4.4 Hyperparameters Sensitivity 5 Conclusion References Modal Interaction-Enhanced Prompt Learning by Transformer Decoder for Vision-Language Models 1 Introduction 2 Methodology 2.1 Text Prompt for Few-Shot Learning 2.2 The Proposed Method 3 Experiments 3.1 Few-Shot Learning 3.2 Domain Generalization 3.3 Ablation Studies 3.4 Further Analysis 4 Conclusions References Unveiling Cybersecurity Threats from Online Chat Groups: A Triple Extraction Approach 1 Introduction 2 Related Work 2.1 Thread Disentanglement 2.2 Dialogue-Level RE 3 Design and Implementation 3.1 Data Collector 3.2 Ontology Construction 3.3 Dialogue Extractor 3.4 Knowledge Triple Extractor 4 Evaluation 4.1 Experiments for Thread Disentanglement 4.2 Experiments for Relation Extraction 5 Conclusion References KSRL: Knowledge Selection Based Reinforcement Learning for Knowledge-Grounded Dialogue 1 Introduction 2 Model 3 Experimental Evaluation 3.1 Quantitative Results 3.2 Qualitative Results 4 Conclusion References Prototype-Augmented Contrastive Learning for Few-Shot Unsupervised Domain Adaptation 1 Introduction 2 Related Work 2.1 Unsupervised Domain Adaptation 2.2 Contrastive Learning 2.3 Contrastive Learning for Unsupervised Domain Adaptation 3 Method 3.1 Two-Stage Prototype Computation for Source Domain 3.2 Prototype Computation for Target Domain 3.3 In-Domain Prototype Contrastive Learning 3.4 Cross-Domain Prototype Contrastive Learning 3.5 PAC Learning for FS-UDA 4 Experiments 4.1 Experimental Settings 4.2 Experimental Results 4.3 Ablation Study 5 Conclusion References Style Augmentation and Domain-Aware Parametric Contrastive Learning for Domain Generalization 1 Introduction 2 Related Work 2.1 Data Augmentation 2.2 Contrastive Learning 3 Method 3.1 Problem Definition 3.2 Overall Framework of the Model 3.3 Style Augmentation 3.4 Domain-Aware Parametric Contrastive Learning 4 Experiments 4.1 Experiment Details 4.2 Comparison with State-of-the-art Methods 4.3 Single-Source Domain Generalization 4.4 Analysis 5 Conclusion References Recent Progress on Text Summarisation Based on BERT and GPT 1 Introduction 2 BERT-Based Model 2.1 Single Document Summarisation 2.2 Multi-Document Summarisation 2.3 Speech Summarisation 2.4 Non-English Text Summarisation 2.5 Hybrid Method 3 GPT-Based Model 3.1 Long Document Summarisation 3.2 Multi-Document Summarisation 3.3 Dialogue Summarisation 3.4 Non-English Text Summarisation 4 BERT and GPT-Based Text Summarisation 4.1 English Text Summarisation 4.2 Non-English Text Summarisation 5 Application 5.1 Legal Domain 5.2 News Text Summarisation 5.3 Healthcare Domain 6 Challenges 7 Conclusion References Ensemble Strategy Based on Deep Reinforcement Learning for Portfolio Optimization 1 Introduction 2 Problem Description 2.1 State Space 2.2 Action Space 2.3 Reward Function 3 The Ensemble Strategy 3.1 Selection of the Base Trading Agent 3.2 Ensemble Strategy 4 Experiments 4.1 Experimental Environment and Datasets 4.2 Evaluation Baselines and Indicators 4.3 Performance Comparisons 5 Conclusion References A Legal Multi-Choice Question Answering Model Based on BERT and Attention 1 Introduction 2 Model Structure 2.1 Dataset of Articles, Questions, and Answer Options 2.2 Article Retriever 2.3 BERT Fine-Tuning 2.4 Information Fusion 2.5 Information Match 2.6 Answer Choose 2.7 Model Training 3 Experiments 3.1 Dataset 3.2 Baselines 3.3 Settings 3.4 Benchmark Experiment 3.5 Ablation Experiment 3.6 Experiments of Article Number 3.7 Benchmark with ChatGPT 4 Relate Work 4.1 Multi-Choice Question Answering 4.2 Legal Question Answering 5 Conclusion References Offline Reinforcement Learning with Diffusion-Based Behavior Cloning Term*-7pt 1 Introduction 2 Related Works 3 Preliminaries 3.1 Offline Reinforcement Learning 3.2 Diffusion Model 4 Method 4.1 The Design of Behavior Cloning Term 4.2 Practical Implementation 5 Experiments 5.1 Performance on D4RL Benchmark 5.2 Ablation Study 6 Conclusion References Evolutionary Verbalizer Search for Prompt-Based Few Shot Text Classification*-7pt 1 Introduction 2 Related Work 2.1 Verbalizer Construction 2.2 Evolutionary Algorithm 3 Evolutionary Verbalizer Search 3.1 Encoding and Decoding 3.2 Population Evaluation 3.3 Verbalizer Generation Strategies 4 Experiments 4.1 Experimental Settings 4.2 Baselines 4.3 Implementation Details 4.4 Results 4.5 Analysis 5 Conclusion References Graph Contrastive Learning Method with Sample Disparity Constraint and Feature Structure Graph for Node Classification 1 Introduction 2 Preliminaries 2.1 Attributed Networks 2.2 Graph Contrastive Learning 3 The Proposed Method 3.1 Feature Structure Graph Construction 3.2 Node Embeddings Generation 3.3 Disparity Constraint and Optimisation of Objective Functions 4 Experiments 4.1 Experimental Setup 4.2 Classification Performance Analysis 4.3 Hyperparameter Analysis 4.4 Convergence Analysis 5 Conclusion References Learning Category Discriminability for Active Domain Adaptation 1 Introduction 2 Related Work 2.1 Active Learning (AL) 2.2 Domain Adaptation (DA) 2.3 Active Domain Adaptation (ADA) 3 Method 3.1 Overall Framework 3.2 Learning Discriminative Features via Co-training Two Classifiers 3.3 Active Query with Task-Specific Classifiers 3.4 Training with a Progressively Augmented Labeled Target Set 4 Experiments 4.1 Datasets and Implementations 4.2 Main Results 4.3 Analysis 5 Conclusion References Multi-level Contrastive Learning for Commonsense Question Answering 1 Introduction 2 Related Work 3 Methods 3.1 Knowledge Integration 3.2 Multi-level Contrastive Learning 3.3 Training Objective 4 Experiments 4.1 Datasets and Baselines 4.2 Implementation Details 4.3 Main Results 4.4 Ablation Study 4.5 Effects of Problems Improvement 4.6 Generality Analysis 4.7 Case Study 5 Conclusion References Efficient Hash Coding for Image Retrieval Based on Improved Center Generation and Contrastive Pre-training Knowledge Model 1 Introduction 2 Related Work 2.1 Deep Hash Coding for Image Retrieval 2.2 Pre-training Model for Knowledge Feature Extraction 3 The Proposed Framework and Method 3.1 Hash Coding Framework for Image Retrieval 3.2 Improved Hash Center Generation Procedure 3.3 Loss Function Designed for Deep Model 3.4 Contrastive Pre-training Knowledge Model 4 Experiments 4.1 Datasets and Baselines 4.2 Evaluation Metric and Implementation Details 4.3 Experimental Results and Analysis 5 Conclusion References Univariate Time Series Forecasting via Interactive Learning 1 Introduction 2 Related Works 2.1 Time Series Forecasting Methods 2.2 Attention-Based Blocks 3 Methodology 3.1 Time Series Decomposilion 3.2 Interactive Temporal-Spatial Attention (ITSA) Block 3.3 Hierarchical Stacking of ITSA 4 Experiments 4.1 Experimental Setup 4.2 Forecasting Results on ETT Dataset 4.3 Dropout Rates Forecasting for MOOC Dataset 4.4 The Effect of Hyperparameters 4.5 Ablation Study 5 Conclusion References Task Inference for Offline Meta Reinforcement Learning via Latent Shared Knowledge 1 Introduction 2 Method 2.1 Problem Formulation 2.2 MeTask: Learning Meta-knowledge from Tasks 3 Experiments 3.1 Experimental Settings 3.2 Results and Analysis 4 Conclusion References A Quantitative Spectra Analysis Framework Combining Mixup and Band Attention for Predicting Soluble Solid Content of Blueberries 1 Introduction 2 Materials and Methods 2.1 Data Preparation and Acquisition 2.2 Spectra Extraction 3 Proposed Network 3.1 Mixup Module 3.2 Band Attention Module 3.3 Regression Network 4 Results and Discussion 4.1 Spectra Profiles 4.2 Dataset Partitioning and Experimental Setting 4.3 Experimental Results 5 Conclusion References Contextualized Hybrid Prompt-Tuning for Generation-Based Event Extraction 1 Introduction 2 Related Work 3 Methodology 3.1 Structure Generation for Event Extraction 3.2 Hybrid Prompt-Tuning 3.3 Training 4 Experiments 4.1 Experimental Settings 4.2 Results in Fully-Supervised Event Extraction 4.3 Results in Low-Resource Event Extraction 4.4 Ablation Study 5 Conclusion and Future Work References udPINNs: An Enhanced PDE Solving Algorithm Incorporating Domain of Dependence Knowledge 1 Introduction 2 Related Works 3 Proposed Methodology 3.1 Shortcomings of Existing Methods 3.2 Unidirectional Information Propagation Mechanism 3.3 Caching Mechanism for Communication Area 4 Experiments and Results 4.1 Ordinary Differential Equation 4.2 Wave Equation 4.3 Heat Equation 4.4 Incompressible Flow Equations 5 Conclusion References Joint Community and Structural Hole Spanner Detection via Graph Contrastive Learning 1 Introduction 2 Related Work 2.1 Structural Hole Spanner Detection 2.2 Community Detection 2.3 Graph Contrastive Learning 3 Preliminary 4 Method 4.1 Augmentation-Free Contrastive Layer 4.2 Trainable Clustering Layer and Modularity Objective 4.3 Joint Optimization 4.4 Structural Hole Spanner Score Function 5 Experiments 5.1 Datasets 5.2 Implementation Details 5.3 Community Detection 5.4 Structural Hole Spanner Detection 5.5 Ablation Study 5.6 Case Study 6 Conclusion References A Reinforcement Learning-Based Approach for Continuous Knowledge Graph Construction 1 Introduction 2 Related Work 3 Methodology 3.1 Overview 3.2 Reward-Based Question Generation 3.3 Question Answering 3.4 Knowledge Graph Updating 4 Experiments and Results 4.1 Dataset 4.2 Automatic Evaluation 4.3 Human Evaluation 5 Conclusion and Future Work References A Multifactorial Evolutionary Algorithm Based on Model Knowledge Transfer 1 Introduction 2 Proposed Algorithm MT-MFEA 2.1 Clustering Model Building Strategy 2.2 Model Knowledge Transfer Strategy 2.3 Local Optimum Escaping Strategy 2.4 Main Structure of MT-MFEA 3 Experimental Studies 3.1 Test Problems 3.2 Experimental Settings 3.3 Experimental Results and Analysis 4 Conclusion References Knowledge Leadership, AI Technology Adoption and Big Data Application Ability 1 Introduction 2 Literatures Review and Research Hypotheses 3 Research Design 3.1 Research Framework 3.2 Measurement 3.3 Sampling 4 Data Analysis 4.1 Reliability and Validity Analysis 4.2 Test of Overall Theoretical Model 4.3 Test of Hypothesis 5 Conclusions and Discussion References RFLSem: A Lightweight Model for Textual Sentiment Analysis 1 Introduction 2 Methodology 2.1 RFLSem Model 2.2 Data Encoder with FGM 2.3 Transformer with Random Seed 2.4 Adding Linear Layer 3 Experiments 3.1 Research Questions 3.2 Subjects Under Study 3.3 Variables 3.4 Parameter Settings 4 Results and Analyses 4.1 RQ1 – Comparison with Typical Sentiment Analysis Techniques 4.2 RQ2 – Comparison with Classical Pre-trained Models 4.3 RQ3 – Comparison with the Optimized Pre-trained Models 4.4 RQ4 – Ablation Study About Optimized Strategies 5 Related Work 6 Conclusion References Author Index