PRICAI 2023: Trends in Artificial Intelligence: 20th Pacific Rim International Conference on Artificial Intelligence, PRICAI 2023, Jakarta, Indonesia, ... II (Lecture Notes in Artificial Intelligence)

✍ Scribed by Fenrong Liu (editor), Arun Anand Sadanandan (editor), Duc Nghia Pham (editor), Petrus Mursanto (editor), Dickson Lukose (editor)

• Publisher: Springer
• Year: 2023
• Tongue: English
• Leaves: 515
• Category: Library

✦ Synopsis

This three-volume set, LNCS 14325-14327 constitutes the thoroughly refereed proceedings of the 20th Pacific Rim Conference on Artificial Intelligence, PRICAI 2023, held in Jakarta, Indonesia, in November 2023.
The 95 full papers and 36 short papers presented in these volumes were carefully reviewed and selected from 422 submissions. PRICAI covers a wide range of topics in the areas of social and economic importance for countries in the Pacific Rim: artificial intelligence, machine learning, natural language processing, knowledge representation and reasoning, planning and scheduling, computer vision, distributed artificial intelligence, search methodologies, etc.

✦ Table of Contents

Preface
Organization
Contents – Part II
Machine Learning/Deep Learning
A Spatial Interpolation Method Based on BP Neural Network with Bellman Equation
1 Introduction
2 Related Work
3 NNRB Method
3.1 BP Neural Network
3.2 Data Gridding
3.3 Bellman Equation Processes Residuals
4 Experimental Results
4.1 Parameters and Settings
4.2 Results of Spatial Interpolation Over Soil Nutrient Datasets
5 Conclusion
References
Attention Auxiliary Supervision for Continuous Sign Language Recognition
1 Introduction
2 Related Works
2.1 Mechanism of Attention
2.2 Auxiliary Supervision
3 Our Approach
3.1 Main Stream Design
3.2 CTC Loss for Alignment
3.3 Knowledge Distillation for CSLR
3.4 Attention Auxiliary Supervision
4 Experiments
4.1 Experimental Setup
4.2 Ablation Studies
4.3 Comparison with State-of-the-Arts
5 Conclusions
References
AutoShape: Automatic Design of Click-Through Rate Prediction Models Using Shapley Value
1 Introduction
2 Related Work
3 Method
3.1 Search Space of AutoShape
3.2 Search Strategy
3.3 Training and Prediction of Supernet
4 Experiments and Results
4.1 Experiment Setting
4.2 Performance Comparison
4.3 Ablation Experiments
5 Conclusion
References
Byzantine-Robust Federated Learning via Server-Side Mixtue of Experts
1 Introduction
2 Related Work
2.1 FedAvg ch4mcmahan2017communication
2.2 Byzantine-Robust Federated Learning
3 Our Approach
3.1 Overview of FLSMoE
3.2 Model of FLSMoE
3.3 FLSMoE Algorithm
4 Experiments
4.1 Settings
4.2 Results
5 Conclusion
References
CDAN: Cost Dependent Deep Abstention Network
1 Introduction
1.1 Kernel Based Approaches for Learning with Abstention
1.2 Neural Network Based Approaches for Learning with Abstention
1.3 Proposed Approach
2 Multiclass Reject Option Classifier
3 Bounded Multiclass Abstention Loss
4 Proposed Approach: Cost Dependent Abstention Network (CDAN)
4.1 CDAN: Input Independent Rejection
4.2 CDAN: Input Dependant Rejection
4.3 CDAN: Input Dependent Rejection with Auxiliary Heads
5 Experiments
5.1 Datasets Used
5.2 Baselines for Large Datasets Experiments:
5.3 Baselines for Small Datasets Experiments:
5.4 Experimental Settings
5.5 Network Architecture and Implementation Details
5.6 Empirical Observations on Small Datasets
5.7 Empirical Observations on Large Datasets
6 Conclusions and Future Research Directions
References
Learning and Regression on the Grassmannian
1 Introduction
2 The Geometric Structure
3 The Regression Model
3.1 Formulation
3.2 Inferring on the Optimal Model
3.3 A Specific Example of the Grassmannian
4 Experimental Results
5 Conclusion
References
Partial Multi-label Learning with a Few Accurately Labeled Data
1 Introduction
2 Related Work
3 Proposed Method
3.1 Problem Setting
3.2 Confidence Estimation with the Noisy Set and the Validation Set
3.3 Learning a Classifier
4 Experiments
4.1 Experiments Setting
4.2 Comparison Results
4.3 Parameter Analysis
5 Conclusion
References
Shareable and Inheritable Incremental Compilation in iOOBN
1 Introduction
2 Background
2.1 Inference
3 SIIC Compilation Algorithm
3.1 Efficiency Analysis
4 Experimental Analysis
5 Conclusions and Future Work
References
A Dynamic Pricing Strategy in Divided Regions for Ride-Hailing
1 Introduction
2 Basic Settings
3 Dynamic Region-Division Based Pricing Strategy
3.1 Dynamic Region-Clustering Algorithm
3.2 Adaptive Multi-Region Dynamic Pricing Algorithm
4 Experimental Analysis
5 Conclusion
References
CANAMRF: An Attention-Based Model for Multimodal Depression Detection
1 Introduction
2 Methodology
2.1 Feature Extractor
2.2 Adaptive Multi-modal Recurrent Fusion
2.3 Hybrid Attention Module
2.4 Training Objective
3 Experiments
3.1 Baselines
3.2 Main Results
4 Conclusion
References
CASSOR: Class-Aware Sample Selection for Ordinal Regression with Noisy Labels
1 Introduction
2 Our Method
2.1 Preliminaries
2.2 Overall Framework
2.3 Class-Aware Sample Selection
2.4 Regularization with Label Ranking
3 Experiments
3.1 Experimental Settings
3.2 Experimental Results
3.3 Ablation Study
4 Conclusion
References
Incomplete Multi-view Weak-Label Learning with Noisy Features and Imbalanced Labels
1 Introduction
2 Methodology
2.1 Auto-weighted Incomplete Multi-view Embedding
2.2 Imbalanced Weak-Label Embedding
2.3 Correlation Modeling by Auto-weighted HSIC
2.4 The Proposed NAIL Method
3 Experiments
3.1 Experimental Settings
3.2 Experimental Results
4 Conclusion
References
Natural Language Processing
A Joint Framework with Audio Generation for Rare Gunshot Event Detection
1 Introduction
2 Related works
3 Methods
3.1 GGD Architecture
3.2 Audio Generation
3.3 Offline Mode for GGD
4 Experiments and Analysis
4.1 Dataset
4.2 Implementation Details
4.3 Experiments Results
5 Conclusion
References
Ancient Chinese Machine Reading Comprehension Exception Question Dataset with a Non-trivial Model
1 Introduction
2 Related Work
2.1 Ancient Chinese Reading Comprehension
2.2 Reading Comprehension via Pre-trained Language Models
3 ACRE
3.1 Task Specification
3.2 Statistics of ACRE
3.3 Data Collecting
3.4 Data Biases and Challenges
4 EVERGREEN
4.1 Problem Formalization
4.2 Overview of Network Architecture
4.3 Formalized Procedure
4.4 Evidence Extraction
4.5 Entire-Text Convolution
5 Experiments
5.1 Experiment Settings
5.2 Model Comparison
5.3 Results Analysis
5.4 Ablation Test
6 Conclusion
References
Chinese Macro Discourse Parsing on Generative Fusion and Distant Supervision
1 Introduction
2 Related Work
3 Discourse Parsing on Generative Fusion
3.1 T5-Based Encoder
3.2 Decoder on Discriminant Model
3.3 Decoder on Generative Model
4 Discourse Parsing on Distant Supervision
5 Experimentation
5.1 Dataset and Experimental Settings
5.2 Experimental Results
5.3 Ablation Analysis
5.4 Analysis on Different Target Sentences
6 Conclusion
References
GHGA-Net: Global Heterogeneous Graph Attention Network for Chinese Short Text Classification
1 Introduction
2 Related Work
3 Proposed Method
3.1 Global Graph Representation
3.2 Hierarchical Graph Attention
3.3 Integrated Heterogeneous Feature Learning
4 Experiments
4.1 Datasets
4.2 Experimental Setup
4.3 Performance Comparison
4.4 Ablation Study
5 Conclusions
References
KSRE-CNER: A Knowledge and Semantic Relation Enhancement Framework for Chinese NER
1 Introduction
2 Related Work
3 Methodology
3.1 Feature Encoding Module
3.2 Graph Module
3.3 Transformer Module
3.4 Fusion and Decoding Module
4 Experiments
4.1 Datasets
4.2 Implementation Details and Evaluation Metrics
4.3 Main Results
4.4 Ablation Study
5 Conclusion
References
Low-Frequency Aware Unsupervised Detection of Dark Jargon Phrases on Social Platforms
1 Introduction
2 Related Work
3 Methodology
3.1 Candidate Dark Jargon Phrases Selection Module
3.2 Phrase-Level Context Representation Generation Module
3.3 Similarity Calculation's Dark Jargon Phrases Detection Module
4 Experiments
4.1 Experiment Setting
4.2 Comparison Experiment
5 Conclusions
References
Multilayer Vision and Language Augmented Transformer for Image Captioning
1 Introduction
2 Method
2.1 Transformer Basic Models
2.2 Visual Contextual Relationship Module
2.3 Attention Enhancement Module
2.4 Objectives
3 Experiments
3.1 Implement Details
3.2 Ablation Studies
3.3 Results and Analysis
4 Conclusions
References
Neural Machine Translation with an Awareness of Semantic Similarity
1 Introduction
2 Related Work
3 Sentence Semantic-Aware Machine Translation
3.1 Problem Definition
3.2 Model Architecture
3.3 Semantic Vector Generator
4 Experiments
4.1 Experimental Settings
4.2 Results
5 Conclusion
References
Optimizing Answer Representation Using Metric Learning for Efficient Short Answer Scoring
1 Introduction
2 Related Work
3 Methodology
3.1 Problem Definition
3.2 Preliminary: Metric Learning
3.3 Defining the Answer Similarity with Score and Rubric
3.4 Downstream Scoring Procedure
4 Experiments
4.1 Datasets
4.2 Metrics
4.3 Baselines
4.4 Implementation Details
5 Scoring Results and Analysis
6 Answer Embedding Space Analysis
7 Conclusion
References
Prompting Generative Language Model with Guiding Augmentation for Aspect Sentiment Triplet Extraction
1 Introduction
2 Our Approach
2.1 Task Definition
2.2 Generative Prompt
2.3 Guiding Augmentation
2.4 Training and Inference
3 Experiments
3.1 Datasets
3.2 Baseline Methods
3.3 Implementation Details
3.4 Main Result
3.5 Ablation Study
3.6 More Analysis
3.7 Case Study
4 Related Work
5 Conclusion
References
Prompting GPT-3.5 for Text-to-SQL with De-semanticization and Skeleton Retrieval
1 Introduction
2 Related Work
3 Methodology
3.1 Question De-semanticization
3.2 LLM-Based Adjustable Prompting
4 Experiment
4.1 Experimental Setup
4.2 Main Results
4.3 Ablation Study
4.4 Case Study
5 Conclusion and Future Work
References
QURG: Question Rewriting Guided Context-Dependent Text-to-SQL Semantic Parsing
1 Introduction
2 Related Work
3 Preliminaries
3.1 Task Formulation
3.2 Relation-Aware Transformer (RAT)
4 Methodology
4.1 Question Rewriting Model
4.2 Rewriting Matrix Construction
4.3 QURG: SQL Parser with Rewriting Matrix
5 Experiments
5.1 Experimental Setup
5.2 Experimental Results
5.3 Ablation Study
6 Conclusions
References
Sarcasm Relation to Time: Sarcasm Detection with Temporal Features and Deep Learning
1 Introduction
2 Related Work
3 Proposed Method
3.1 Data Acquisition
3.2 Data Preprocessing
3.3 Sarcasm Detection
3.4 Temporality Detection
3.5 Lexicons
3.6 Experimental Results
4 Results and Discussion
4.1 Classification Results
4.2 Logistic Regression vs. Existing Works
4.3 Performance Comparison Among Feature Sets
5 Conclusion Remarks and Future Works
References
Self-agreement: A Framework for Fine-Tuning Language Models to Find Agreement Among Diverse Opinions
1 Introduction
2 Related Work
3 Problem and Method
3.1 Consensus-Building Problem
3.2 Automatic Opinion-Agreement Data Generation
3.3 Scoring Agreement Candidates
3.4 Fine-Tuning the Language Model
4 Evaluation
4.1 Evaluation Setting
4.2 Evaluation Results
4.3 Fine-Tuning Analysis
5 Conclusion
References
Self-SLP: Community Detection Algorithm in Dynamic Networks Based on Self-paced and Spreading Label Propagation
1 Introduction
2 Background and Related Work
2.1 Community Detection in Dynamic Network
2.2 Self-paced Learning with Diversity
2.3 Prior Research
3 The Proposed Method
3.1 Soft Spreading Strategy
3.2 Self-paced Propagation Learning with Diversity
3.3 Belonging Verification
3.4 Time Complexity
4 Experiment
4.1 Baseline Methods
4.2 Quality Measurement
4.3 Experiments on Real-World Networks
4.4 Experiments on Synthetic Networks
5 Conclusion
References
Word Segmentation of Hiragana Sentences Using Hiragana BERT
1 Introduction
2 Related Work
3 Proposed Method
3.1 Unigram BERT Word Segmentation System
3.2 Bigram BERT Word Segmentation System
4 Data
4.1 Pre-training Data from Wikipedia
4.2 Word Segmentation Data for Hiragana Sentences from Wikipedia
4.3 Word Segmentation Data for Hiragana Sentences from BCCWJ
4.4 Data for the Hiragana KyTea Word Segmentation System
5 Experiment
5.1 Experiment 1: Fine-Tuning with BCCWJ
5.2 Experiment 2: Fine-Tuning with Wikipedia
5.3 Evaluation Methods
6 Results
7 Discussion
8 Conclusions
References
An Empirical Study on Context Length for Open-Domain Dialog Generation
1 Introduction
2 Experimental Setup
3 Results and Discussion
3.1 Does Longer Context Help Model Training?
3.2 Is It Necessary to Change the Training Context Length When Dealing with Dialogs of Different Context Lengths?
3.3 Do Different Samples Have the Same Preference for Context Length?
4 Conclusion
References
COVER: A Heuristic Greedy Adversarial Attack on Prompt-Based Learning in Language Models
1 Introduction
2 The Proposed Method
3 Experiments
4 Conclusion
References
Few-Shot Table-to-Text Generation with Structural Bias Attention
1 Introduction
2 Methodology
2.1 Preliminaries
2.2 Structural Bias Framework (SF)
2.3 Table-to-Text Generation Models
3 Experiments and Results
3.1 Datasets and Baselines
3.2 Results
3.3 Case Study
4 Conclusion
References
Generalized Knowledge Distillation for Topic Models
1 Introduction
2 Overview of Neural Topic Models
3 Methodology
4 Experiments and Results
5 Conclusions
References
Improving Speaker Recognition by Time-Frequency Domain Feature Enhanced Method
1 Introduction
2 Method
2.1 Time-Frequency Domain Feature Enhanced Method
2.2 Deep Speaker Network
3 Experimental Setup
4 Results and Analysis
5 Conclusion
References
Leveraging Dual Encoder Models for Complex Question Answering over Knowledge Bases
1 Introduction
2 Related Work
3 An Approach
3.1 Query Graph Generation
3.2 Query Graphs Ranking
3.3 Optimization
4 Experimental Studies
4.1 Settings
4.2 Results and Analysis
5 Conclusion
References
Unsupervised Contrastive Learning of Sentence Embeddings Through Optimized Sample Construction and Knowledge Distillation
1 Introduction
2 Distillation-Based Unsupervised Sentence Representation Learning
2.1 Model Architecture
2.2 Data Augmentation Strategy
2.3 Knowledge Distillation
3 Experiments
3.1 Datasets
3.2 Results and Discussion
3.3 Ablation Study
4 Conclusion
References
Optimization
Automatically Choosing Selection Operator Based on Semantic Information in Evolutionary Feature Construction
1 Introduction
2 Related Work
2.1 Multi-armed Bandit
2.2 Automatic Operator Selection
3 The New Algorithm
3.1 Model Representation
3.2 Algorithm Framework
3.3 Selection Operators
3.4 Dynamic Multi-armed Bandit
4 Experimental Settings
4.1 Experimental Dataset
4.2 Parameter Settings
4.3 Evaluation Protocol
4.4 Baseline Algorithms
5 Experimental Results
5.1 Comparison Between Selection Operators
5.2 Comparison of Credit Assignment Strategies
6 Conclusions
References
Evolving a Better Scheduler for Diffusion Models
1 Introduction
2 Related Works
3 Methodology
4 Experiments
5 Discussions
5.1 Optimal start and end
5.2 Choices of Fitness Functions
5.3 Population Size
6 Conclusion and Future Work
References
Investigating the Existence of Holey Latin Squares via Satisfiability Testing
1 Introduction
2 Preliminaries
3 Modeling
3.1 Standardization
3.2 SAT Encoding
4 Symmetry Breaking
5 Benchmarks and Results
6 Experiments
6.1 Comparison of At-Most-One Encodings
6.2 Comparison with Other Solvers
6.3 Effect of Symmetry Breaking
7 Concluding Remarks
References
Leiden Fitness-Based Genetic Algorithm with Niching for Community Detection in Large Social Networks
1 Introduction
2 Related Works
3 Problem Formulation
4 Scalability of Leiden-Based Genetic Algorithm (LGA)
5 Proposed Algorithm
5.1 Overall Algorithm Design
5.2 Construction of the Individual
5.3 Population Initialization
5.4 Fitness Evaluation
5.5 Niching Method
5.6 Genetic Operators
6 Experiment and Analysis
6.1 Benchmark Networks
6.2 Baseline Algorithms
6.3 Parameter Settings
6.4 Experiment Result Analysis
7 Conclusions
References
Non-revisiting Stochastic Search for Automatic Graph Learning
1 Introduction
2 Background
2.1 Bilevel Optimization
2.2 Automatic Graph Learning
3 Algorithm
3.1 Tree Building Process
3.2 Non-revisiting Process
4 Experiments
4.1 Experimental Setup
4.2 Experimental Results
5 Conclusion
References
Detecting AI Planning Modelling Mistakes – Potential Errors and Benchmark Domains
1 Introduction
2 AI Planning Formalism
3 Potential Errors in Planning Domains
3.1 Syntax Errors
3.2 Semantic Errors
4 Evaluation of Existing Parsers
4.1 Results
5 Conclusion
References
Responsible AI/Explainable AI
Decision Tree Clustering for Time Series Data: An Approach for Enhanced Interpretability and Efficiency
1 Introduction
2 Related Work
3 Method
3.1 Branch Features and Time Series Features
3.2 Surrogate Silhouette Coefficient
3.3 Algorithm for Decision Tree Construction
4 Numerical Experiments
4.1 Comparison of Computational Time
4.2 Comparison Based on Distributions
4.3 Application for Time Series Data
4.4 Accuracy of Time Series Prediction Using Clustering Results
5 Conclusion
References
The Ethical Evaluation Method of Algorithmic Behavior Based on Computational Experiments
1 Introduction
2 Related Work
2.1 Ethical Research on Algorithmic Behavior
2.2 Computational Experiments
3 The Ethical Evaluation Method of Algorithmic Behavior
3.1 Overall Framework
3.2 Artificial Society
3.3 Recommendation Algorithms and Metrics
4 Experiment
4.1 Experiment Setup
4.2 Analysis of Experimental Results
5 Conclusion
References
Metrics for Evaluating Actionability in Explainable AI
1 Introduction
2 Study Design and Methodology
2.1 Scenarios
2.2 Actionability Metrics
2.3 Online Study
3 Results and Discussion
3.1 Experimental Condition One and Two
3.2 Experimental Condition Three
4 Conclusion
References
Author Index