Ground truth? Concept-based communities versus the external classification of physics manuscripts

EPJ Data Science

Table 2 Basic characteristics of the datasets

	N	V	\(\boldsymbol{V_{\mathrm{gen}}}\)	〈 k 〉	\(\boldsymbol{L^{\mathrm {in}}_{\mathrm{idf}}}\)	\(\boldsymbol{L_{\mathrm{idf}}}\)	\(\boldsymbol{L^{\mathrm{in}}_{\mathrm {bp}}}\)	\(\boldsymbol{L_{\mathrm{bp}}}\)
arxivPhys2013	36,386	12,200	347	37	5.9 × 10⁸	3.3 × 10⁸	2.1 × 10⁶	1.3 × 10⁶
arxivPhys2014	41,848	12,728	344	38	7.8 × 10⁸	4.5 × 10⁸	2.5 × 10⁶	1.6 × 10⁶

Total number of articles (N), total number of identified concepts (V) and the number of generic ones (\(V_{\mathrm{gen}}\)) among them; 〈k〉 gives the average number of non-generic concepts within arbitrary chosen article. The number of links in a unipartite network provided that the generic concepts are included (\(L^{\mathrm {in}}_{\mathrm{idf}}\)) or excluded (\(L_{\mathrm{idf}}\)) is two orders of magnitude larger than the corresponding number of links in bipartite networks (\(L^{\mathrm{in}}_{\mathrm{bp}}\) and \(L_{\mathrm{bp}}\), respectively). This results in significant differences in computational resources needed to perform community detection analysis.