基于概念的视频检索中概念语义匹配算法研究

Application of Semantic Similarity Algorithm in Concept-based Video Retrieval

为了实现基于概念视频检索中从底层内容到查询的语义贯通,应用基于WordNet词典的语义相似度算法,通过对三种不同原理的算法对比应用,得出基于信息量算法在本应用中更有优势,语义匹配可以提高检索精度,最优映射数目为2至3个,以及在目前发展水平下,映射到合适的概念比检测器精度更合适四个重要结论.

In order to run semantic linking through the bottom content to users＇ query in concept-based video retrieval, we apply semantic similarity algorithm based on WordNet to video retrieval. By comparing the effects of three different principles algorithm, we get four important conclusions. Algorithm based on information entropy is better for use here than others, semantic matching can improve retrieval precision, the optimal number of maps is 2 to 3, and in the current development level, mapping to the appropriate detector is important than the detector accuracy.