Social Networks (SNs) connect nodes from different geographical areas, keeping users updated about current affairs through message sharing. Natural calamities or deliberately imposed actions can cause Internet disconnections between geographical areas. This results in a SN partition which leads to communication loss between nodes inside the partitioned area. In this paper, we propose an extended Multi-Dimensional Routing (eMDR) algorithm using Greedy routing, which considers multiple attributes for routing. It improves the communication efficiency inside partitioned SNs. The performance of the proposed algorithm is validated by considering three dimensions/attributes, viz., social interest, geographical location and time-zones of social nodes on both real and synthetic SN datasets. The results of topological and routing probabilities for Chord and novel Social Interest Overlay networks, show considerable improvement in communication inside partitioned SNs.