Self-learning Knowledge Models

Semantic knowledge models capture the expert and background knowledge about a particular domain and involved stakeholders, e.g. user profiles, medical knowledge, context, etc. This knowledge is then exploited by applications to offer personalized & context-aware intelligent services to the end-users. However, knowledge constantly evolves and new insights should be able to be incorporated in the knowledge model. Therefore, the main research objective of this track is to devise algorithms, platforms and techniques that enable self-learning and evolving knowledge models that accurately reflect a domain.

Specifically, our research efforts focus on the following three topics:

  • Design of data mining techniques that allow to process and exploit the captured semantic data in order to discover new semantic links and concepts, e.g., using inductive logic programming or association rule mining.
  • Design of algorithms and techniques that allow to exploit the semantics captured in knowledge models in order to improve the accuracy and performance of machine learning techniques, e.g., to boost learning speed and generalization capability, to enhance feature selection & extraction techniques, to improve model parameter initialization, etc.
  • Design of platform and techniques that allow that support evolving knowledge models, supporting multiple views on a domain through probabilistic logics and reasoning.
  • A participatory ontology engineering methodology that allows to accurately and incrementally construct knowledge models together with domain experts without requiring medical knowledge from them.

Staff

Filip De Turck, Femke Ongenae, Femke De Backere, Erik Mannens, Ruben Verborgh

Researchers

Pieter Bonte, Gilles Vandewiele, Rein Houthooft, Laurens De Vocht, Anastasia Dimou, Pieter Heyvaert

Projects

  • ICON AORTA: Adaptive Optimization for Resource and Task Allocation in Hospitals
  • ICON ACCIO:Ambient aware provisioning of Continuous Care for Intra-muros Organizations
  • ICON – WONDER: interventions for Wandering and Other behavioral disturbaNces of persons with DEmentia in nursing homes by personalized Robot interactions
  • Bilateral ESS: Environmental Sensing Study

Key publications

A self-learning platform able to 1) enhance the data analytics performed on data through semantics, 2) enhance the knowledge model by incorporating semantic patterns discovered in the data.
A self-learning platform able to 1) enhance the data analytics performed on data through semantics, 2) enhance the knowledge model by incorporating semantic patterns discovered in the data.

 

The different tracks on which research is performed into incorporating semantic background knowledge to steer white-box machine learning techniques and achieve a higher accuracy and a more interpretable model.
The different tracks on which research is performed into incorporating semantic background knowledge to steer white-box machine learning techniques and achieve a higher accuracy and a more interpretable model.

 

Impressions of workshops performed as part of the Participatory ontology engineering methodology where (extensions of existing) knowledge models are constructed in close collaboration with the domain experts without requiring technical or semantic knowledge from them.
Impressions of workshops performed as part of the Participatory ontology engineering methodology where (extensions of existing) knowledge models are constructed in close collaboration with the domain experts without requiring technical or semantic knowledge from them.