论文部分内容阅读
准确识别作者研究内容的相似度,是探测学科知识结构和挖掘潜在合作关系的重要基础工作,也是近年来图书情报学的研究热点。现有的相似度计算方法大都依赖于属性的直接关联,忽略属性间的间接关联。提出一种新的基于作者关键词网络的作者相似度计算方法,通过向量空间模型计算出关键词之间的关联度,再利用图结构相似度算法P-Rank挖掘出作者间的间接关联关系。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和向量空间模型算法,该方法可以明显地提高作者相似度计算的准确性。
Accurately identifying the similarity of the author’s research content is an important basic work in exploring the knowledge structure of the subject and in exploring the potential cooperation. It is also a hot spot in the study of library information science in recent years. Most of the existing similarity calculation methods rely on the direct association of attributes, ignoring the indirect association between attributes. This paper proposes a new method for calculating the similarity of authors based on the author keyword network. The relevance between keywords is calculated by vector space model, and then the indirect relation between authors is extracted by using the P-Rank algorithm of graph structure similarity. Preliminary experiments show that this method can effectively identify the similarity between authors, which can obviously improve the accuracy of authors’ similarity calculation compared with the traditional keyword coupling and vector space model algorithms.