基于文献内聚度的非相关文献知识发现排序方法研究

Research on the Literature Cohesion-based Ranking Method of Disjoint Literature-based Discovery

在对现有非相关文献知识发现中间集排序方法进行分析的基础上,以共现理论为基础,以主题关联度为着眼点,提出基于文献内聚度加权的B排序方法。并以Swanson的早期发现之一为基础,考察经文献内聚度加权和逆文献频率加权两种方法排序筛选后B的范围以及目标关联词和目标关联对的出现情况,以此作为评价其对B影响的依据。结果表明基于文献内聚度加权法能显著提高B的质量,从而提高发现效率。

Based on the analysis of the existing ranking methods of B collection of disjoint literature- based discovery, this paper proposes literature cohesion - based ranking method according to the co - occurrence theory and the subject relative degree. Then, an experiment is conducted comparing with one of Swanson＇ s former discoveries. The size of B and the occurrence of the target terms and target relations are explored to evaluate the effect on B of the two methods including the literature cohesion - based weight and the inverse document frequency - based weight. The results of the experiment indicate that the literature cohesion - based ranking method can improve the quality of B and enhance efficiency of discovery accordingly.