We propose an innovative approach for academic recommendation, which integrates the topic component into the graph-based recommendation. Specifically, based on the topic component extracted from the text data on academic corpus, we propose a User Behavior Graph (UBG) and the User Academic Network (UAN). UBG is a bridge to fuse topic information and UAN. A personalized PageRank (PPR) algorithm is used on UBG for recommendation. Experimental results show that our method for academic recommendation outperforms the content-based recommendation.