Labeling-Based Knowledge Construction For Real-World Event Understanding

A sensor networked environment is capable of observing real-world phenomena as sensor readings, however, there are difficulties in understanding real-world 'events' such as activities of daily life occurring in the environment. We propose an incremental model for constructing real-world knowledge to allow us to understand real-world events. The central plank of the proposed model is the labeling practice. In the model, both the ontology of real-world events and the implementation of a sensor system arc simultaneously improved based on a manually labeled event corpus. A labeling tool is developed in accordance with the model and event vocabularies are evaluated in a practical labeling experiment.