Improving retrieval of plane geometry figure with learning to rank

摘要

Educational images are increasingly becoming available online, but an effective method to search for such images is nonexistent, particularly for graph-based digital resources. This paper focuses on plane geometry figure (PGF) retrieval with ranking optimization to retrieve relevant digital geometry materials. A learning to rank model is introduced to rearrange the unsatisfactory order of highly similar PGFs in retrieval results. Moreover, to enhance the retrieval accuracy and efficiency, we perform feature selection for ranking according to the quality and redundancy of several specific types of PGF features. We perform retrieval experiments and evaluations on two PGF datasets, and results show that our PGF retrieval method improves figure retrieval accuracy better than existing methods. (C) 2016 Elsevier B.V. All rights reserved.