SNAP : An Action-Based Ontology for E-commerce Reasoning

This paper presents SNAP, an e-commerce ontology developed for automated recommendation systems for the domains of banking, insurance, and telephony. This ontology is distinguished by the fact that its core concepts come from commonsense theories of action; its support of derived relations, which rely on composition of basic relations; and its close connection to a sound (with respect to first-order logic) and efficient reasoning mechanism. We present the basic concepts and relations of the ontology, and explain how derived relations are built using the construction operators for regular expressions. We describe the knowledge structure and reasoning mechanism that are used on instantiations of the ontology to produce customer recommendations. We then discuss how this ontology has been used in practical applications.