In order to improve efficiency, we first choose the entities with closest label to construct the mapping space, then we propagate the mapping pair with higher similarity, namely compose mapping pair in entities related with them and add to mapping space.