MDATA: A New Knowledge Representation Model: Theory, Methods and Applications

✍ Scribed by Yan Jia, Zhaoquan Gu, Aiping Li

• Publisher: Springer
• Year: 2021
• Tongue: English
• Leaves: 265
• Series: Lecture Notes in Computer Science 12647
• Edition: 1
• Category: Library

✦ Synopsis

Knowledge representation is an important task in understanding how humans think and learn. Although many representation models or cognitive models have been proposed, such as expert systems or knowledge graphs, they cannot represent procedural knowledge, i.e., dynamic knowledge, in an efficient way.

This book introduces a new knowledge representation model called MDATA (Multi-dimensional Data Association and inTelligent Analysis). By modifying the representation of entities and relations in knowledge graphs, dynamic knowledge can be efficiently described with temporal and spatial characteristics. The MDATA model can be regarded as a high-level temporal and spatial knowledge graph model, which has strong capabilities for knowledge representation. This book introduces some key technologies in the MDATA model, such as entity recognition, relation extraction, entity alignment, and knowledge reasoning with spatiotemporal factors. The MDATA model can be applied in many critical applications and this book introduces some typical examples, such as network attack detection, social network analysis, and epidemic assessment.

The MDATA model should be of interest to readers from many research fields, such as database, cyberspace security, and social network, as the need for the knowledge representation arises naturally in many practical scenarios.

✦ Table of Contents

Preface
Organization
Contents
Introduction to the MDATA Model
1 Background
2 Knowledge Representation Models
2.1 Symbolic Logic Model
2.2 Semantic Network
2.3 Expert System
2.4 Semantic Web
2.5 Knowledge Graph
3 Cognitive Models
3.1 ACT-R Model
3.2 IBLT Model
3.3 SOAR Model
4 What is the MDATA Model
4.1 Overview of the Model
4.2 Knowledge Representation
4.3 Knowledge Acquisition
4.4 Knowledge Usage
4.5 Comparison with Other Models
5 Typical Applications of the MDATA Model
5.1 Network Attack Detection
5.2 Social Network Analysis
5.3 Personnel Association Analysis Against Epidemic
6 Chapter Summary
References
The Framework of the MDATA Computing Model
1 Introduction
1.1 The Development of Cyberspace
1.2 The Requirements for the Characteristics of Big Data
1.3 ``Information Island'' and Protection of Intellectual Property Rights
2 Definition of Fog-Cloud Computing
3 The Architecture of Fog-Cloud Computing
3.1 Cloud Knowledge Actors
3.2 Middle Layer Knowledge Actors
3.3 Fog End Knowledge Actors
3.4 Multi-knowledge-actor Collaborative Computing
4 Building MDATA Model on Fog-Cloud Computing Framework
4.1 Distributed Knowledge Representation of MDATA
4.2 Distributed Knowledge Acquisition of MDATA
4.3 Distributed Knowledge Usage of MDATA
5 Typical Applications
5.1 Network Attack Detection
5.2 Epidemic Analysis
6 Chapter Summary
References
Spatiotemporal Data Cleaning and Knowledge Fusion
1 Introduction
2 Spatiotemporal Data Cleaning
2.1 Background
2.2 Problem Definition
2.3 Related Methods
2.4 Discussions on Spatiotemporal Data Cleaning for MDATA
3 Knowledge Fusion
3.1 Background
3.2 Problem Definition
3.3 Related Methods
3.4 Discussions on Knowledge Fusion for MDATA
4 Chapter Summary
References
Chinese Named Entity Recognition: Applications and Challenges
1 Introduction
2 Background
2.1 Development of NER Method
2.2 Applications of Chinese NER
3 Problem Definition
3.1 The Difference Between Chinese NER and NER
3.2 Challenge in Chinese NER
4 Related Methods
4.1 Word Based Chinese NER
4.2 Character Based Chinese NER
4.3 Lexicon Based Chinese NER
5 Challenges and Ideas When Combining with MDATA
6 Chapter Summary
References
Joint Extraction of Entities and Relations: An Advanced BERT-based Decomposition Method
1 Introduction
2 Related Work
3 Preliminary
3.1 Task Description
3.2 Problem Analysis
4 Method
4.1 The Tagging Scheme
4.2 The End-to-End Model
5 Experiments
5.1 Dataset
5.2 Experiment Settings
5.3 Experiment Results
6 Connections with the MDATA Model
7 Chapter Summary
References
Entity Alignment: Optimization by Seed Selection
1 Introduction
2 Background
2.1 Crowdsourcing and Rules Based Alignment
2.2 Similarity Calculation
2.3 Entity Embedding
3 Related Work
4 Iterative Entity Alignment Method
4.1 Problem Definition
4.2 Node Centrality
4.3 Iterative Entity Alignment
5 Experiment
5.1 Datasets
5.2 Experiment Setting
5.3 Comparison of Experimental Results
6 Challenges of Entity Alignment in the MDATA Model
7 Chapter Summary
References
Knowledge Extraction: Automatic Classification of Matching Rules
1 Introduction
2 Background
3 Related Methods of Word2vec
4 Method
4.1 Problem Definition
4.2 Automatic Classification of Matching Rules Using Word2vec
5 Experiment
6 Opportunities and Challenges of Knowledge Extraction in the MDATA Model
7 Chapter Summary
References
Network Embedding Attack: An Euclidean Distance Based Method
1 Introduction
2 Background
3 Problem
4 Related Methods
4.1 Network Embedding
4.2 Adversarial Attacks
5 Our Method
5.1 Euclidean Distance Attack
5.2 Experiments
6 Connection with the MDATA Model
7 Chapter Summary
References
Few-Shot Knowledge Reasoning: An Attention Mechanism Based Method
1 Introduction
2 Background
2.1 Related Work
2.2 Knowledge Reasoning Model
3 Preliminary
3.1 Problem Definition
3.2 Model Description
4 Knowledge Reasoning
4.1 Embedding Model
4.2 Reasoning Model
4.3 Model Training
5 Experiment
5.1 DataSet
5.2 Experiment Design
6 Chapter Summary
References
Applications of Knowledge Representation Learning
1 Introduction
2 Approaches of Knowledge Representation Learning
2.1 Distance Model
2.2 Semantic Matching Model
2.3 Bilinear Model
2.4 Neural Network Model
2.5 Embedding with Additional Information
3 Evaluation Tasks and Datasets of Knowledge Representation Learning
3.1 Evaluation Tasks and Metrics
3.2 Benchmark Datasets
4 Applications in Downstream Tasks
4.1 Recommendation System
4.2 Relation Extraction
4.3 Question Answering
5 Association with MDATA
6 Chapter Summary
References
Detection and Defense Methods of Cyber Attacks
1 Introduction
2 APT Background Knowledge
2.1 Definition of APT Attacks
2.2 Basic Characteristics of APT Attacks
2.3 Trends in APT Attacks
3 Detection Methods of APT Attacks
3.1 Social Engineering Based Detection Methods
3.2 Anomalous Flow Based Detection Methods
3.3 Machine Learning Based Detection Methods
4 Defense Strategies Against APT Attacks
4.1 APT Attack Defense Strategy Based on Limited Resources
4.2 APT Attack Defense Strategy Based on Information Flow Tracking
4.3 Defense Strategy for Cloud-Based APT Attacks
5 Relations with MDATA Model
6 Chapter Summary
References
A Distributed Framework for APT Attack Analysis
1 Introduction
2 Background
3 Attack Analysis Framework
3.1 Cyber Security Knowledge Graph Construction
3.2 Threat Element Association Statistics
3.3 Threat Element Association Statistics
4 Experimental Results
4.1 Verify the Validity of Self-defined Reasoning Rules
4.2 Verify the Feasibility of the Analysis Framework
5 Chapter Summary
References
Social Unrest Events Prediction by Contextual Gated Graph Convolutional Networks
1 Introduction
2 Related Work
2.1 Graph Convolution Network
2.2 Event Prediction Based on Twitter or GDELT Datasets
2.3 Squeeze and Excitation (SE) Module
3 Methods
3.1 Graph Construction
3.2 Graph Convolutional Network Layer
3.3 Contextual Gated Layer
3.4 Output Layer
3.5 Construction of the Evolution Graph
4 Experiment
4.1 Datasets
4.2 Data Preprocessing
4.3 Prediction Performance
4.4 The Evolution Graph
5 Challenges
6 Chapter Summary
References
Information Cascading in Social Networks
1 Introduction
2 Background
2.1 Social Network Graph Data
2.2 Traditional SNICA Methods
2.3 Embedding Methods of Social Network Graph Data
2.4 Deep Learning Models in SNICA
2.5 The Advantages of Deep Learning in SNICA
3 A General Framework for Deep Learning Methods in SNICA
3.1 Preprocessing of Social Network Data
3.2 Data Embedding
3.3 Selection of Deep Learning Models
3.4 Task Classification
4 Application Classification in SNICA
4.1 User Behavior Analysis
4.2 Information Cascade Prediction
4.3 Rumor Detection
4.4 Analysis of Social Network Event
5 Future Direction
5.1 SNICA Data Perspective
5.2 SNICA Deep Learning Model Perspective
6 Chapter Summary
References
Author Index