中文摘要：

现有的相似度计算方法大都依赖于作者间的直接关联，忽略了间接关联。文章提出一种新的基于SimRank的作者相似度计算方法，充分考虑作者关键词二分图网络的整体结构特性，利用图结构相似度算法挖掘出作者间以及词汇间的潜在关联关系。初步实验表明该方法能够有效地识别作者之间的相似度，相比于传统的关键词耦合，该方法可以明显提高作者相似度计算的准确性。

英文摘要：

Existing similarity calculation methods mostly depend on the direct correlation between authors as well as ignore the indirect correlation. This paper proposes a new calculation method of author similarity based on SimRank. The method gives much at- tention to the overall structure characteristics of the author keyword bipartite graph network, and uses graph structure similarity algo- rithm to mine the potential relationship between authors and vocabularies. The preliminary experiment demonstrates that the method can effectively identify the similarity between authors, and this method can obviously improve the accuracy of author similarity cal- culation with the comparison of the traditional keywords coupling.