Residential and commercial buildings account for more than 74% of total annual electricity consumption in the United States. Studies have shown that occupants' awareness of their behaviors in consuming electricity encourages them to change their unsustainable behaviors and improves the sustainable ones. As behaviors impact the ways that daily activities are performed, in order to develop a personalized appliance level model of an occupant's behavior, precise activity recognition is required. In this paper, we introduce a novel framework to allocate personalized appliance-level disaggregated electricity consumption to daily activities. In our framework, using ontology based approach, the input appliance usage data is first separated into categories of non-overlapping activity events. The separated data sets are then segmented to detect activity segments, which are next mapped into activity classes using a trained classification model. To evaluate the performance of our presented framework, an experimental validation was carried out in three test bed apartment units. Results of validation showed a total F-measure value of 0.97 for segmentation and an average accuracy of 93.41% for activity recognition. Following the activity recognition, the approximate electricity consumption associated with the recognized activities was estimated and the results of each test bed unit were compared with the others.