Actionable Knowledge Graphs - how daily activity applications can benefit from embodied web knowledge

This Thesis proposes a five-step methodology for creating actionable knowledge graphs that follows existing knowledge engineering standards but links object knowledge to environment and action knowledge to enable various applications in daily environments, on different agents. The methodology is exemplary applied in two scenarios with different foci to create a product knowledge graph and a food cutting knowledge graph. The product knowledge graph aims at enabling omni-channel applications in unknown environments. It therefore contains product-related knowledge that is used by different agents such as smartphone, smart glass and robot, which aim at providing shopping assistance in a retail store. In order to provide user assistance like routing a customer to a searched product on different devices such as robot or smartphone, this scenario focuses on accessing relevant Web knowledge about products in a retail store that is linked to precise, reliable and agent-independent environment information. The food cutting knowledge graph aims at enabling robots to execute task variations of cutting actions. Here, the idea is to access Web knowledge to enable a robot to autonomously perform a range of cutting tasks. Therefore, this scenario focuses on how object information can influence action execution, how the needed knowledge can be acquired from the Web and how it can be modelled in a knowledge graph in such a way that a robot can use it to execute tasks. The methodology is validated by showcasing various applications that are enabled by the two exemplary knowledge graphs. The applications range from smartphone applications for shopping assistance that highlight interesting product features or route to a searched product over smart glass applications like shopping assistance and a recipe application to robot applications for shopping assistance and execution of cutting task variations on different fruits and vegetables.