Paper Abstract
"The study of commonsense reasoning and robotic agents in home environments is trending considering artificial intelligence. Handling the contexts of human daily living is crucial for the system to achieve, for example, the detection of risky situations in a care facility and long-term natural dialogue with other individuals. This study aims to construct an ontology to represent daily living activities by reusing the International Classification of Functioning, Disability and Health (ICF) categories. We extracted relevant categories from the ICF, introduced new classes, and defined an “is-a” hierarchy to provide a common vocabulary for the annotation of the context of multiple sensor data obtained in home environments. As a result, the constructed ontology consists of 284 classes. The evaluation of its coverage is performed using the following two resources related to commonsense reasoning: ATlas Of MachIne Commonsense (ATOMIC) and STAIR Actions captions. We sampled 100 sentences from each resource and described the context of the samples using the constructed ontology, which resulted in 63% and 84% of each sample being covered by our ontology. This demonstrates that our ontology has the competency to describe the context in a daily living environment. In addition, we found the types of contexts that are lacking in the ontology via the evaluation."

The full paper is available through SpringerLink, provided that you have proper access rights.