Data-knowledge Integration Research Team
The Data Knowledge Integration (DKI) research team specializes in
goal-oriented knowledge graphs and ontology development, context-aware dialogue
modeling for AI-based interaction management.
AI technology is rapidly progressing into society and the people's
every-day life, ranging from child-care, care-services and safety management to
well-being, healthy living, and active aging. It is thus important
that AI can be used to realize solutions and services that support
people's daily activities and enable friendly, efficient and
trustworthy interactions between humans and between humans and
autonomous applications in smart environments. However, such
interactions require understanding of the people's intentions,
motions, and interactions, as well as knowledge about actions,
objects, and their relations and roles in the human activities.
While neural network based AI achieved recognition of objects and motions in many applications in some extent, integration of knowledge required to interpret human daily activity into data driven AI still remains to be an ambitious challenge.
To develop AI systems for real-world situations, it is necessary
to recognize and interpret human daily activities,
integrate them to structured knowledge (manuals, safety guidelines, law, etc.)
not embodied in the observed data and generate intelligent reactions.
X to Knowledge Graph technology integrates structured knowledge into multimodal and sensor observation data to provide a computational framework, which enables us to develop AI technologies that exploits hybrid modeling
techniques (rule-based and neural models) for AI-based interaction management.
Using such technologies, the DKI team is working
on uncovering the semantics of human behaviour, and developing AI
technology that can recognize and support various episodes of daily
activities by integrating observation data and knowledge into a
context-aware dialogue model that allows natural interaction in a
variety of cases where it is necessary for humans to have dialogue
interactions with other agents and with the smart environment.