DBIS

Overview

Ontologies are formal, machine-readable knowledge representations that provide shared vocabularies for a domain. They are a key enabler of semantic data integration, allowing heterogeneous data sources to be interpreted in a common, unambiguous way, and as such form the backbone of knowledge graphs and interoperable data infrastructures in fields such as materials science and biomedicine. However, community-maintained ontologies such as EMMO can span thousands of concepts, making it impractical to import them in full: doing so is computationally expensive, introduces irrelevant concepts that degrade retrieval quality, and overwhelms domain experts tasked with validating mappings. This thesis investigates how compact, application-specific ontology modules can be extracted automatically from large ontologies, preserving the semantic coherence needed for downstream tasks while drastically reducing complexity.

Research Questions

• Which ontology reduction strategies exist? How can they be structured?
• How can extracted ontology modules be evaluated, both intrinsically (coverage, logical consistency) and extrinsically (downstream annotation or mapping quality)?
• Which ontology reduction strategies yield the best performance with regard to efficiency and output quality?
• Given a use case, what can an effective pipeline look like? Which challenges arise during its implementation and by which steps can it be streamlined?

Methodology

The research will involve researching, classifying, and comparatively evaluating ontology reduction strategies, as well as designing and implementing a specific pipeline for ontology reduction following a single or a combination of multiple strategies. Quantitative and qualitative metrics will be applied to evaluate the efficiency and integration quality.

Tasks

• Comprehensive literature review and analysis on ontology modularization and module extraction techniques
• Implementation of a practicable ontology reduction pipeline based on one or a combination of multiple strategies
• Experimental evaluation with quantitative and qualitative analysis
• The solution should be integrated into the existing tool KONDA to evaluate the effect on semantic annotation quality.

Initial Literature

• OntoPath: a query language for retrieving ontology fragments (2007)
• COMET: an ontology extraction tool based on a hybrid modularization approach (2021)
• Ontology Module Extraction for Ontology Reuse (2007)
• Ontology View: a new sub-ontology extraction method (2014)
• Task Oriented Evaluation of Module Extraction Techniques (2009)
• Relevant Subgraph Extraction from Random Walks in a Graph (2006)

In case you are interested in this thesis, please write an email to the thesis advisor with your CV and transcript of records.