Intuitive Network Applications: Learning for Personalized Converged Services Involving Social Networks

The convergence of the wireline telecom, wireless telecom, and internet networks and the services they provide offers tremendous opportunities in services personalization. We distinguish between two broad categories of personalization systems: recommendation systems, such as used in advertising, and life-style assisting systems, which attempt to customize or specialize services to an individual’s needs, preferences, and habits. The Privacy-Conscious Personalization (PCP) framework, developed previously at Bell Labs, uses a high-speed rules engine to enable rich life-style assisting personalization. During network-hosted information sharing and call processing, the PCP framework can be used to interpret a combination of incoming requests, user data, and user preferences in order to provide contextaware, requester-targeted, and preferences-driven responses to those requests (e.g., deciding whether to share a user’s location with a given requester, what to show as the enduser’s availability to a given requester, where to forward an incoming call). This paper describes key aspects of a new initiative at Bell Labs, called Intuitive Network Applications (INA), which aims to combine human factors and automated learning techniques, in order to gather the user data and preferences needed for PCP-enabled personalization, with minimal disruption to the user. A particular focus of the paper is on life-style assisting capabilities for applications that involve the interaction of an end-user with her social network, i.e., family, friends, colleagues, customers, etc. The paper describes (i) key requirements, (ii) a high-level architectural framework, and (iii) some specific directions currently under exploration for filling out the framework.