4. PAL Ontology
owl file
Links to the owl files will be provided per sub-ontology (see the corresponding pages).
The overall PAL ontology can be found via the following link.
Authors
Michael van Bekkum, Bernd Kiefer, Hans-Ulrich Krieger, Rifca Peters and Willeke van Vught
Short description of ontology
The Federated Ontology of the PAL Project,. The PAL ontology is the overall ontology that facilitates the links between all the different ontologies. It specifies equivalent classes over the ontologies and specifies some cross relations over ontologies if this was easier for implementation. The PAL ontologies loads all the different sub ontologies and therefore this is the ontology that is actually used in the PAL system.
The entire ontology in the PAL project is constructed by integrating separate ontologies, linking them by means of a top-level ontology. These separate models function as high-level building blocks for smaller, more specific areas of interest (frames) . In PAL, we have reused existing ontologies to cover the various frames wherever possible. Although the frames of interest for PAL are typically generic in nature, pre-existing models for these frames may differ (slightly) in scope and/or intention and may thus be a partial fit to the intended scope of the frame in the context of PAL. Whereas e.g. self-management activities of diabetes are a relevant topic, the entire professional medical diagnosis and treatment model of diabetes is out of scope. We have adapted some of the existing models by either extending them with additional concepts or by taking a profile (part) from the model whenever there are details/concepts in the model that are irrelevant to the scope of PAL. An example of reuse is displayed in the adoption of the well-known ontology for task world modelsin the frame for tasks/goals and learning objectives.
Overall, the PAL ontology currently consists of eight sub-ontologies, seven of which are "truly" independent and capture "really" different knowledge. One further ontology brings them together through the use of hand-written interface axioms, employing axiom constructors such as rdfs:subClassOf and owl:equivalentProperty, or by posing domain and range restrictions on certain under-specified properties. The TBox (in short, the concepts) and RBox (in short, the hierarchy of the concepts) of the PAL domain stays constant, i.e., will not change over time. Only relation instances from the ABox (in short, the value of a concept) might undergo a temporal change, e.g., the weight of a child at certain times, but, e.g., not the birthdate. The independence of the subontology has as advantage that it provides clean sub-ontologies which can be reusable in other projects and/or domains. Besides that, this structure has a practical advantage that different project partners can work on the ontologies simultaneously, without interfering with each other.
Related work
van Bekkum, M. A., Krieger, H. U., Neerincx, M. A., Kaptein, F., Kiefer, B., Peters, R., ... & Folmer, E. (2016, January). Ontology engineering for the design and implementation of personal pervasive lifestyle support. In SEMANTiCS (Posters, Demos, SuCCESS).
Neerincx, M. A., Kaptein, F., Van Bekkum, M. A., Krieger, H. U., Kiefer, B., Peters, R., ... & Sapelli, M. (2016). Ontologies for social, cognitive and affective agent-based support of child’s diabetes self-management. Artificial Intelligence for Diabetes, 35.
Ontology design
The ontology loads all PAL sub-ontologies and enables specifications and/or relations where needed.
Evaluation and results
The PAL ontology is implemented in the PAL system and enabled the ontologies and modules to exchange information. In particular, the ontology engineering approach proved to be helpful for the incremental development of the PAL knowledge base and PAL’s dialogues and user interfaces. First, the creation of the ontology made tacit knowledge of the health-care professionals explicit in a formal (logically correct) model that is interpretable by the relevant human stakeholders and the PAL system. It resulted in an extendable set of self-management objectives (focusing on learning), with a coherent and concise structure. From this structure, new content was generated, such as quiz questions, break&sort tasks, new memory “card set”, educational videos and timeline tasks, which are all tied together per child consistently (i.e., supporting transparent personalization). Second, via the ontologies, social and cognitive theories have been integrated into the PAL system in a transparent and verifiable way (e.g., affect, memory, agreement and explanation). Third, the content with its structure of the dashboards and dialogues are based on the ontology. Finally, it should be noted that ontology engineering is an iterative process. Draft ontologies for Feedback and Explanations have been developed for PAL, which are being completed and validated in current research.