National and Kapodistrian University of Athens
From knowledge graphs to insights: cognitive and geospatial analysis and visualization of linked data
Abstract
dc:descriptionIn recent years, the rapid growth of Semantic Web technologies has brought forth unprecedented opportunities for enriched data representation, Knowledge Graph construction, and services that harness the power or linked data. However, the challenge remains in making linked data accessible to users across diverse domains. There are two main barriers that users face when they design applications that are powered by linked data. First, these applications necessitate the integration of interlinked subsets from different Knowledge Graphs to perform analytical operations. Second, such applications must feature intuitive interfaces that empower end-users to effectively navigate linked data structures and extract meaningful insights. This dissertation addresses these challenges through two principal contributions: (i) Sextant, a web-based system for the visualization and exploration of linked geospatial data, that models maps as RDF resources and provides tools to interact with the data to extract insights and (ii) MindCraft, a system that implements ``Semantic Selective Attention'', a cognitively-inspired methodology for subgraph extraction from Knowledge Graphs, that assists users in designing personalized interlinked subgraphs to fuel their applications. Collectively, these systems mitigate the inherent barriers of linked data consumption, enabling application development and knowledge discovery. Sextant builds on the foundation of Semantic Web technologies to provide a versatile system for dynamic map composition, where geospatial datasets, enriched with temporal and thematic dimensions, can be visualized and queried. Empirical evaluation utilizing the System Usability Scale yielded an "Excellent" rating, indicating that the system is well-aligned with the user needs and successfully mitigated the complexities of linked geospatial data analysis and visualization. By abstracting the complexity of SPARQL querying and RDF data modeling, Sextant has been deployed across multiple European projects to make linked data accessible to practitioners in fields like environmental monitoring, emergency response, and urban analytics. MindCraft’s methodology utilizes cognitive constructs such as Episodic Memory, Working Memory, and Semantic Memory in subgraph extraction. Drawing inspiration from Baddeley’s Working Memory Model, MindCraft aims to improve the usability and interpretability of extracted data, while aligning closely with human memory and cognition processes. The system guides the user through the ontology schema of the Knowledge Graph, to select classes and properties that are used to build their subgraphs. It supports subgraph interlinking using entity resolution, and integrates with Large Language Models to further augment a subgraph. An empirical assessment using the System Usability Scale classified MindCraft's usability as "Excellent", indicating that the system effectively supports users in constructing and extracting semantically coherent subgraphs. To further investigate how the proposed methodology shapes the user experience during subgraph extraction, we conducted a Thematic Analysis that demonstrates that the ``Semantic Selective Attention'' paradigm aligns closely with users’ natural cognitive processes, and shows how MindCraft emerges as a system for subgraph extraction that introduces new forms of engagement that simultaneously empower users while exposing AI-limitations related to trust and transparency.Together, MindCraft and Sextant establish a unified, two-stage framework for Linked Data analysis. In the first stage, MindCraft allows users to navigate large Knowledge Graphs, apply ``Semantic Selective Attention'', integrate prior knowledge, and extract semantically coherent and interconnected subgraphs. In the second stage, Sextant receives these extracted datasets and translates them into interactive, multi-layered cartographic visualizations. The development and evaluation of both systems demonstrate that the Semantic Web can become significantly more accessible without compromising the power and expressiveness of its tools. This is achieved through suitably designed interfaces that simplify access by masking complex technical challenges, and empower both beginners and domain experts in these technologies to utilize linked data for analysis and extraction of meaningful insights.
Degree
thesis:*- Grantor dc:publisher
- National and Kapodistrian University of Athens
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Stamoulis, George
- Σταμούλης, Γεώργιος
Subjects
dc:subject × 22- Διασυνδεδεμένα δεδομένα
- Γεωχωρικά δεδομένα
- Οπτικοποίηση δεδομένων
- Γράφοι γνώσης
- Εξαγωγή υπογραφήματος
- Οντολογίες
- Ανθρώπινη σκέψη
- Linked data
- Geospatial data
- Data visualization
- Knowledge graphs
- Subgraph extraction
- Ontologies
- Human cognition
- Επιστήμη Ηλεκτρονικών Υπολογιστών και Πληροφορική
- Φυσικές Επιστήμες
- Διάδραση ανθρώπου-υπολογιστή
- Πληροφοριακά συστήματα
- Computer and Information Sciences
- Natural Sciences
- Human-Computer Interaction
- Information Systems
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/62598