Geospatial link discovery with human in the loop / Abdullah Fathi Ahmed Ahmed ; 1. Reviewer: Prof. Dr. Axel-Cyrille Ngonga Ngomo, Department of Computer Science (Paderborn University) ; 2. Reviewer: Prof. Dr. Alsayed Algergawy, Department of Computer Science (University of Passau) ; Supervisors. Prof. Dr. Axel-Cyrille Ngonga Ngomo and Dr. Mohamed Ahmed Sherif. Paderborn, 2024
Inhalt
- Titlepage
- Abstract
- Acknowledgements
- Contents
- List of Selected Publications
- I Preliminaries
- 1 Introduction
- 1.1 Motivation
- 1.2 Research Gaps and Contributions
- 1.2.1 Research Gap 1: Geospatial Link Discovery over Knowledge Graphs at Scale
- 1.2.2 Research Gap 2: The Absence of Holistic Models for Linked Open Data
- 1.2.3 Research Gap 3: Explainable Link Discovery
- 1.3 Thesis Outline
- 2 Notation
- 2.1 Linked Data
- 2.2 Knowledge Graphs
- 2.3 Ontology Matching
- 2.4 Link Discovery
- 2.5 Knowledge Graph Fusion
- 2.6 Link Prediction
- 2.7 Well-Known Text Format for Geometries
- 2.8 Topological Relations
- 3 Related Work
- 3.1 Link Discovery
- 3.2 Geospatial Link Discovery over Knowledge Graphs
- 3.2.1 Orchid
- 3.2.2 Silk
- 3.2.3 RADON
- 3.2.4 GIA.nt
- 3.2.5 DORIC
- 3.2.6 Strabon
- 3.2.7 Geo-L
- 3.2.8 MaskLink
- 3.2.9 Line Simplification
- 3.2.10 Point-Sets Distance Measures
- 3.3 Holistic Models for Linked Open Data
- 3.4 Explainable Link Discovery
- II Improving Link Discovery Scalability
- 4 LineSimp: A Line Simplification Approach For Link Discovery over Geospatial Knowledge Graphs
- 5 RADON2: An Intersection Matrix Approach For Link Discovery over Geospatial Knowledge Graphs
- 6 Cobalt: A Content-Based Similarity Approach For Link Discovery over Geospatial Knowledge Graphs
- 6.1 Motivation
- 6.2 Approach
- 6.2.1 R-tree Indexing
- 6.2.2 Querying R-trees
- 6.2.3 Building R-trees
- 6.2.4 Content Measures for Topological Relations
- 6.2.5 Combining R-tree Indexing and Content Measures
- 6.2.6 Indexing Both Datasets
- 6.2.7 Splitting Polygons to Gain Accuracy
- 6.3 Evaluation and Results
- 7 Nellie: Never-Ending Linking for Linked Open Data
- III Link Discovery Explainability And Human In The Loop
- 8 Multilingual Verbalization and Summarization for Explainable Link Discovery
- 8.1 Motivation
- 8.2 Template-Based LS Verbalization Approach
- 8.3 Neural-Based LS Verbalization Approach
- 8.4 Selectivity-Based LS Summarization Approach
- 8.5 Evaluation
- 9 Explainable Integration of Knowledge Graphs Using Large Language Models
- 9.1 Motivation
- 9.2 Approach
- 9.2.1 Rule-Based Verbalizer
- 9.2.2 Standard Encoder-Decoder Architectures
- 9.2.3 Few-shot Learning Using the T5 model
- 9.3 Evaluation
- 10 Conclusion and Future Work
- Bibliography
- Declaration
- Declaration