Rank Learning Model of Cross-media Retrieval based on Structured SVM
Ying Jie Xia, Xiao Tingting · 2015
The diversity of Internet information increases the users’ retrieval requirements for the multimedia data across different modals. Compared with existing one-across-one retrieval approaches to get results in only one media modal, we focus on the one-across-all retrieval model to acquire related results from multiple media modals. Ranking is crucial for the quality of information retrieval. A rank learning model for cross-media retrieval is proposed, it takes advantage of listwise rank learning method, utilizes structured SVM framework construct optimization with ranking criteria to learn relevance mapping functions. Application prototype shows that this rank learning method is effective to gain relevant cross-media retrieval results.