Abstract: With the rapid development of Internet technology,the presented forms of network information have exten-ded from simple text to images,voice,video and other multimedia expression.In the field of multimedia information retrieval,the traditional methods often represent all of the media mode in the same feature space model.Existing methods take either one-to-one paired data or uni-directional ranking examples.In this paper,we considered learning bi-directionalranking examples in the cross-media retrieval.By analyzing the experimental results basing on the Wikipedia dataset,it is demonstrated better performance of the proposed method.

Key words: Cross-media representation,Bi-directional learning ranking,Latent space,Similarity measurement