Research on Author's Semantic Similarity Based on Collaborative Network

Abstract

The community is not only one kind of widely existing organization in networks, but also contains distinct information about topics. The present paper is to define a metric between community members, weigh their semantic similarity, and finally make use of the metric to find more about the Web. In order to achieve this, the usual practice is to establish a "plane" adjacency matrix according to the citation relationship among all community members. However, it is easy to trigger the problem of topic drift. To overcome this weak point, the present paper puts forward firstly the strategy of establishing a three-order adjacency tensor on the 3-dimensional relationship between the seed document, simple document and the author. Secondly, the adjacency tensor is decomposed to obtain the principal component in each dimension. Thirdly, the semantic similarity between authors is defined. The experiment makes it clear that the semantic similarity between the author and people of importance tends to be stable under a particular circumstance.