Co-word analysis method based on meta-path of subject knowledge network

We propose a method of co-word analysis based on the subject knowledge network meta-path to overcome limitations with the current co-word analysis method. First, we construct a subject knowledge network to find the word-to-word meta-path. Second, we use the HeteSim algorithm to calculate the semantic relevance between words based on each meta-path. Then, through matrix operations, standardization, and matrix fusion, we construct a word-to-word semantic relevance matrix (WSRM). We conduct an empirical evaluation to test the proposed method. The results indicate that the WSRM formed by this method is superior to the word-to-word similarity matrix used in traditional co-word analysis in terms of both macro-evaluation indicators (viz., network density, network centralization, network average degree, and cohesive subgroups) and micro-evaluation indicators (viz., core-periphery class, point centrality, and cluster analysis). The method overcomes limitations to the traditional co-word analysis method, and combines multiple semantic relations between words, to reflect the relationship between words more realistically.