Extraction of commonsense knowledge for "bring something" robotic service at home

Sharing commonsense knowledge is one key to realize symbiosis between human and robot. However, as human knowledge is so complicated, it is difficult to extract appropriate information for robot. This paper presents our approach to extract the core part from Basic-level Knowledge Network (BKN) to be used for intuitive "Bring something" robotic service. A two-step filtering method was implemented based on the weighting mechanism of BKN. The first step used one weighting parameter to filter the most irrelevant data. The second step combined all weighting parameters from BKN with user response data from a questionnaire to estimate commoness of object-activity connections. Result showed that the proposed approach is efficient to extract the most common activities related to each object in BKN, therefore to help robot understands human intention and provide intuitive service.