This paper proposes a social recommendation algorithm for use in a research social network environment. The social recommendation algorithm proposed combines the concepts of a relationship ontology and item-based collaborative filtering (CF). While the network setup in social networking sites can accurately reflect the social landscape of its users, it is much harder to detect the importance or strength of any one link. We therefore propose an extension to our recommendation algorithm which makes use of the idea of co-presence communities to increase the relevance of the recommendations. A copresence community can be detected from with data collected from Bluetooth-enabled mobiles. Detection of a copresence community can help determine the nature and importance of the social links between participating members.