Extraction of Relevant Axioms Using Ontology Embedding

Supervisor: Jieying Chen (j.chen2@vu.nl)


As the complexity and volume of ontologies grow, ensuring efficient access to the embedded knowledge emerges as a pressing challenge. Many real-world ontologies have grown to sizes that are cumbersome for users to navigate, making comprehension and reasoning a formidable task. The purpose of the thesis is to extract the most relevant axioms w.r.t. users’ interests. Through this approach, we aim to enhance usability and understanding, making vast knowledge maps more digestible for human users.


  1. Investigate the current methodologies used for ontology axiom extraction and the theoretical foundations of various embedding techniques, emphasizing those applicable to ontologies.
  2. Develop a systematic approach to convert axioms in ontologies to continuous vector representations.
  3. Design metrics or methods based on the embedded axioms to quantify and rank their relevance concerning specific criteria or domain requirements.
  4. Compare the embedded-based axiom extraction method against traditional methods, measuring precision, recall, and ranking accuracy for top-k axiom extraction.
  5. Application Use-Cases: Explore real-world scenarios or case studies where extracting the top-k axioms would be beneficial, showcasing the utility and efficacy of the embedding-based extraction approach.
  6. Based on initial results, fine-tune the embedding and ranking methods to improve extraction accuracy and relevance.