基于改进Fish-search算法的多媒体检索

Multimedia Retrieval Based on Improved Fish-search Algorithm

针对标准Fish算法的链接相关度区分度低、搜索宽度不灵活、主题描述不准确的缺点,提出一种改进的算法。在计算链接相关度时,引入网页分块的思想以及增量反馈和自适应机制。在计算搜索宽度时,提出一个启发性规则来动态改变其大小。实验结果表明,改进算法的查准率和查全率分别提高了118%和24.16%。

Since the defect of the standard Fish algorithm that the differentiation of link-relevance is low and the searching width lacks flexibility and topic description is not accurate, this paper proposes an improved algorithm. When computing the link relevance, it introduces the webpage-partition thought and the mechanism of increment feedback and self-adapting. When computing the searching width, it proposes an enlightening rule to change its value dynamically. Experimental results show that the improved algorithm increases the precision ratio and recall ratio by 118% and 24.16% respectively.