摘要
传统文献检索大多按照被引次数、发表时间、搜索词出现频次等条件之一对结果进行排序,角度单一且忽略了文献相互引用带来的价值流动,往往会出现部分文献排名过高或过低的现象。为此,很多国内外学者提出将PageRank算法应用到文献检索中,并取得了一定程度的改进,但是忽略了一些特殊情况,如文献使用价值可能会随时间的推移而产生衰退,还有一些发表时间较短的文献被引次数为零,如何去评估它的价值等。文章针对这些问题,提出了一种多维检索排序法,综合考虑各种因素带来的影响,并引入文献活跃度的概念,以加权的方式将文献价值量化。实验证明,多维检索排序法比传统文献检索排序法效果更好,而且由权值迭代所带来的额外的计算量均为离线完成,在提高准确率的同时也很好地保持了检索的效率。
Most of the traditional literature retrievals sort the results under one of the conditions of cited frequency, publication time or frequency of the searched words. This method always uses a single angle that ignores the value flow of mutually referred articles and this leads to a phenomenon that some literature gets a too high or too low rank. For this reason, many scholars at home and abroad apply the PageRank algorithm to literature retrieval and some improvements have been made, however they ignore some special circumstances, for example, the value of literature may decline over time, and articles with short publication time have no cited record, so we cannot evaluate their value. To solve these problems, a kind of multidimensional retrieval ordering method is proposed in this paper, which gives a comprehensive consideration to all the influence factors, involves the concept of literature activity and quantifies the value of literatures in weighted manner. Experiments show that the proposed retrieval has a better performance than traditional document retrieval, and the extra amount of calculation caused by weight iteration is done offline in order to improve the accuracy and at the same time to maintain the efficiency of the retrieval.
出处
《情报理论与实践》
CSSCI
北大核心
2016年第11期126-130,144,共6页
Information Studies:Theory & Application
关键词
文献检索
多维检索排序
PAGERANK算法
文献活跃度
document retrieval
multidimensional retrieval ordering
PageRank algorithm
literature activity