Computer Applications: 38th CCF Conference of Computer Applications, CCF NCCA 2023, Suzhou, China, July 16–20, 2023, Proceedings

Preface
Organization
Contents – Part I
Contents – Part II
Artificial Intelligence and Applications
PRiskLoc: An Enhanced Multi-dimensional Root Cause Localization Algorithm Aided by a Fine-Grained Evaluation System
1 Introduction
2 Background
2.1 Definition
2.2 Deviation Score
2.3 Related Work
3 Evaluation
3.1 Evaluation System
3.2 Dataset Generation
3.3 Evaluation Metrics
3.4 Results and Discussion
4 PRiskloc
4.1 Inefficiency of 2-Way Partitioning Scheme
4.2 Partition by KDE
4.3 Evaluation of PRiskLoc
5 Conclusion
A Dataset Details
B Result Details
References
Blockchain Enhanced Whole-Life-Cycle Engineering Cost Management
1 Introduction
2 Review of the Research
2.1 Current Situation of Applying Blockchain Technology to the Engineering Cost Management Field
2.2 Whole-Life-Cycle Cost Management and BIM Technology
3 New Trends of Blockchain Technology in Engineering Cost Management
3.1 Data Elements Drive the Digital Transformation and Upgrading of Engineering Cost Management
3.2 The Whole Process of Engineering Cost Database Design Based on Data Mining
4 Application of BIM Technology in the Whole-Life-Cycle Cost Management of Emerging Urban Subway Engineering
5 Conclusion
Bibliography
A Review of Point Cloud 3D Object Detection Methods Based on Deep Learning
1 Introduction
2 Relationship Between Deep Learning and Point Cloud
3 Point Cloud Processing Method
3.1 Based on Graph Convolution
3.2 Based on Original Point Cloud
3.3 Fusion Processing Algorithm
4 Generalize
5 Prospect
References
A Review of Intelligent Opponent Modelling Research for Air Combat Simulation Training
1 Introduction
2 Background and Requirements
3 Key Technologies for Virtual Opponent Modelling
3.1 High-Performance Computing Technology
3.2 Virtual Opponent Modelling Based on Artificial Intelligence Technology
3.3 Virtual Opponent Modelling Technology for Live-Virtual-Constructive (LVC) Training
3.4 Virtual Opponent Anti-deception Technology
3.5 Intelligent Evaluation Technology
4 Conclusion
References
Design of a Blind Guidance System Based on RealSense and the Improved YOLOv5 Algorithm
1 Introduction
2 Methodology
2.1 The YOLOv5-CM Improved Algorithm
2.2 Ground Plane Segmentation
3 Results and Discussion
3.1 Image Data Acquisition
3.2 Model Evaluation Index
3.3 Comparative Experiment
3.4 Ablation Experiment
4 Conclusion
References
AHM: A Novel Model for Mining Academic Hot Spots Based on a Scientific Knowledge Graph
1 Introduction
2 Related Work
3 Methodology
3.1 Data Collection
3.2 Data Preprocessing
3.3 Keyword Extraction
3.4 AHM Model
4 Experimenta. The Results and Discussion
4.1 Data Description
4.2 Model Measurement Metrics
4.3 Baselines
4.4 Extraction of Academic Hot Spots
4.5 Evolutionary Analysis
5 Conclusion
References
How to Define a Multi-modal Knowledge Graph?
1 Introduction
2 Definition of KG
2.1 Description of the Problem
2.2 Inquiry into the Problem
2.3 Knowledge Base, Ontology and RDF
2.4 Conclusion
3 Exploring the Concept of MMKG
3.1 Multimodality of Knowledge Graphs
3.2 Comparison of KG and MMKG
4 Construction of MMKG
4.1 Two Different Ways to Build MMKG
4.2 Building Sample MMKG in the Medical Field
5 Summary
References
Research on Feature Fusion Methods for Multimodal Medical Data
1 Introduction
1.1 A Subsection Sample
2 Background Knowledge
2.1 TransR
2.2 TransE
3 Knowledge Graph Construction
4 Image Classification and Fusion Mode Based on Knowledge Atlas
4.1 Comparison of Optimization Plans
4.2 Gambling Tree
5 Experiment and Analysis
5.1 Evaluating Indicator
5.2 Performance Comparison Experiment of Fusion Pattern Classification Model
5.3 Ablation Experiment
6 Conclusion and Future Work
References
Incorporating Syntactic Knowledge and Position Information for Aspect-Based Sentiment Analysis
1 Introduction
2 Models
2.1 Overview
2.2 Input Layer
2.3 Syntax Fusion Encoder
2.4 Aspect-Context
2.5 Output Layer
3 Experiments
3.1 Dataset
3.2 Implementation Details
3.3 Compared Models
3.4 Main Result
3.5 Analysis
3.6 Effect of the Bi-directional Attention Mechanism
4 Conclusion
References
Integrating Background and General Knowledge for Dialogue Generation
1 Introduction
2 Model
2.1 Background Context Encoder
2.2 Emotional Context Graph and Graph Encoder
2.3 Knowledge Selection
2.4 Response Decoder
3 Experiments
3.1 Implementation Details
3.2 Datasets
3.3 Evaluation Metrics
3.4 Results
3.5 Ablation Study
4 Conclusion
References
Fine-Grained Style Control in VITS-Based Text-to-Speech Synthesis
1 Introduction
2 Method
2.1 VITS
2.2 Training and Inference Procedures
2.3 Emotion Feature Extraction
2.4 Emotion-Text Encoder
3 Experiment Setup
3.1 Comparing Methods
3.2 Subjective Evaluation
3.3 Discussion on the Impact of Transformer on the Model
4 Conclusions
References
Text Sentiment Analysis Based on a Dynamic Pruning Capsule Network
1 Introduction
2 Research Status
3 DP-CAPS-VW-BIGRU Model
3.1 Input Layer
3.2 Word Embedding Layer
3.3 Capsule Network
4 Experiment Settings
4.1 Data Set
4.2 Data Preprocessing
4.3 Parameter Settings
4.4 Evaluating Indicator
5 Experimental Results and Discussion
5.1 Experiment 1
5.2 Experiment 2
5.3 Experiment 3
5.4 Result Analysis
6 Conclusion
References
Research on Ceramic E-commerce Product Review Analysis Based on Semantic Analysis
1 Introduction
2 Semantic Emotion Analysis of the Commentary Text
3 Linear Fitting was Used to Fit the Emotional Value of Commodity Characteristics
4 Conclusion
References
Research on Clarification Question Recognition and Generation in Intelligent Q&A
1 Introduction
2 Design of Clarification Process Framework
2.1 Task 1: Clarification Question Detection
2.2 Task 2: Clarification Question Generation
2.3 Task 3: Entity Prediction
2.4 Overall Process
3 The Model of Clarification Question Detection and Entity Prediction
3.1 Disentangled Attention
3.2 Enhanced Mask Decoder
4 The Model of Clarification Question Generation
4.1 Attribute Tree and Semantic Understanding Module
4.2 ByT5 Model
4.3 Q&A Term Template
5 Experiments
5.1 Dataset
5.2 Evaluation Indicators and Benchmark Models
5.3 Experimental Environment
5.4 Results and Discussion
6 Conclusions and Future Work
Appendix
References
Dictionary-Assisted Chinese Nested Named Entity Recognition
1 Introduction
2 Related Work
2.1 Improved Sequence Labeling Framework
2.2 Hypergraph-Based Approaches
2.3 Parsing Tree-Based Methods
2.4 Span-Based Methods
3 Model
3.1 Import Dictionary Information
3.2 Bi-affine Decoder
3.3 CNN on Score Matrix
3.4 Loss Function
3.5 Entity Decoding
4 Experimental Analysis
4.1 Dataset
4.2 Experimental Settings
4.3 Analysis of Results
5 Ablation Study
6 Conclusion
References
Cce-SpERT: A Span-Based Pre-trained Transformer Model for Chinese Civil Engineering Information Joint Extraction
1 Introduction
2 Data
2.1 Datasets
2.2 The Categories of Entities and Relations
3 Model
3.1 Preprocessing Module
3.2 Transformer Module
3.3 Classification Module
4 Experimental Results
4.1 Pre-training
4.2 Training Parameters
4.3 Training Result of NER with Different Parameters
4.4 NER Results
4.5 RE Result
4.6 Knowledge Graph
5 Conclusions
References
A BMRC Algorithm Based on Knowledge Enhancement and Case Regularization for Aspect Emotion Triplet Extraction
1 Introduction
2 Related Work
2.1 Fine Grained Emotional Analysis
2.2 Machine Reading Comprehension
3 Methodology
3.1 BMRC
4 Model Algorithm
5 Experiment
6 Results
7 Ablation Experiments
8 Conclusion
References
A Method for Extracting Clinical Diagnosis and Treatment Knowledge for Traditional Chinese Medicine Literature
1 Introduction
2 Method
2.1 Technical Process
2.2 Research Objects
2.3 Knowledge Base of TCM Diagnosis and Treatment
2.4 Information Extraction Models
2.5 Data Analysis Methods
3 Experiment
3.1 Data Acquisition
3.2 Data Annotation
3.3 Information Extraction
3.4 Data Analysis
4 Conclusion
References
Application and Optimization of Centroid Algorithm in Indoor Positioning
1 Introduction
2 Method
2.1 Signal Transmission Model
2.2 Principle of the Center-of-Mass Algorithm
2.3 The Proposed Six-Point Centre-of-Mass Algorithm
3 Experiment Result
4 Conclusion
References
MEAHO: Membrane Evolutionary Algorithm for Hyperparameter Optimization of Deep Convolutional Neural Networks
1 Introduction
2 Membrane Evolutionary Algorithm Framework for Hyperparameter Optimization
2.1 Hyperparameter Optimization Problem for Deep Convolutional Neural Networks
2.2 Definition of P System
2.3 Space Exploration
2.4 Membrane Evolutionary Algorithm Framework (MEAF)
3 Hyperparameter Optimization Based on Membrane Evolutionary Algorithm (MEAHO)
4 Experimental Results
4.1 Data Sets
4.2 Evaluation Metrics and Baselines
4.3 Setting of Experimental Parameters
4.4 Results and Analysis
4.5 Self-contrast Experiment
4.6 Structural Validation
5 Discussion
5.1 Membrane Algorithm for Space Exploration
5.2 Evaluation of Hyperparameter Effect
5.3 Mechanism of Membrane Evolution
References
Automatic Detection and Inspection Robot of All-Weather Urban Comprehensive Corridor Based on Voiceprint Detection
1 Introduction
2 Inspection Robot System Design
3 Voiceprint Detection
4 Key Technology of Voiceprint Recognition
5 Experimental Report
6 Conclusion
References
Author Index