Discovering　social　communities　or　social　circles　from　social　networks　is　interesting　and　important　for　many　applications　like　business　advertisement,　social　recommendation　and　collaborative　office.　In　this　paper,　by　integrating　grey　relational　analysis　with　the　label　propagation　algorithm　and　the　parallel　framework,　a　new　parallel　algorithm　for　detecting　overlapping　communities　is　proposed.　The　similarity　of　the　vertices　is　measured　by　the　grey　relational　degree　and　the　parallel　computation　primitives　are　employed　to　propagate　the　labels　in　parallel.　The　experiments　on　both　the　artificial　and　real-world　networks　demonstrate　that　the　new　algorithm　is　effective　in　detecting　overlapping　social　communities.