Phase 3-Prop
Prog. PhD STI
Adm. - Grad. 2016 - 
Dir.; Codir. Stéphane Gagnon

Business Technology Management Ontology and Knowledge Recommendation

Iyer, Shankar

Knowledge recommendation for e-learning contents (e.g., academic articles) is useful to help professionals learn task-specific and role-related practices. Knowledge Graphs can represent domain expertise and its relationship with general vocabularies, as well as professional roles and tasks. They can be modeled using ontologies, which can be processed using query and inference databases, and with Deep Learning for graph embeddings.

Using an Action Design Research methodology (ADR), we propose to develop, validate, test, and implement an Ontology Web Language (OWL) version of the Business Technology Management Body of Knowledge (BTM BOK). We will implement this ontology for the purpose of recommending relevant scientific literature to professional end-users, based on their learning requirements.

Our methodology relies on formal ontology development methods, using such tools as Stanford Protégé and Stardog for ontology design and reasoning. We will work in partnership with a research team involving a network of contributors, who develop the BTM BOK assets using the Eclipse Process Framework (EPF).

The research process will include: (1) import BTM BOK assets published in XMI by EPF, and convert them into RDF-OWL; (2) manually validate the ontology and its integrity, and add 1000+ concepts not yet present in the BTM BOK; (3) align the BTM ontology with the YAGO4 ontology, mostly using manual but also automated ontology alignment tools; (4) build a small web app to represent the ontology to the network of contributors and involve them in the collective process of ontology editing and finalization; (5) build a small web app to allow contributors to make annotation of scientific articles abstracts (at sentence and word level) to indicate how BTM BOK ontology concepts rely to each; (6) use this "gold standard" to test the BTM ontology, by attempting to infer through Stardog reasoning whether each abstract line is properly detected to be the specific ontology concepts as prescribed by end users; (7) analyze results based on F-measure and Matthews Correlation Coefficient (MCC) to evaluate the quality of our ontology inference capabilities; (8) interpret the results as to the necessary improvements in later iterations, and identify a work plan for the community to continue maintain the ontology.


Go to top