Cooccurrence relation extraction of abundant tags is very significant for information retrieve and tag recommendation. This paper presents a parallel extraction method of tag cooccurrence relation based on cloud computing. At last, two experiments show that the parallel extraction method can improve the performance of retrievals greatly and have good scalability. The research may play a role in the facilitative effect on the expanding application of social tagging system with abundant, multiplex, complex annotated data.