Beyond description. Comment on "Approaching human language with complex networks" by Cong & Liu

In their historical overview, Cong & Liu highlight Sausurre as the father of modern linguistics [1]. They apparently miss G. K. Zipf as a pioneer of the view of language as a complex system. His idea of a balance between unification and diversification forces in the organization of natural systems, e.g., vocabularies [2], can be seen as a precursor of the view of complexity as a balance between order (unification) and disorder (diversification) near the edge of chaos [3]. Although not mentioned by Cong & Liu somewhere else, trade-offs between hearer and speaker needs are very important in Zipf’s view, which has inspired research on the optimal networks mapping words into meanings [4, 5, 6]. Quantitative linguists regard G. K. Zipf as the funder of modern quantitative linguistics [7], a discipline where statistics plays a central role as in network science. Interestingly, that centrality of statistics is missing Saussure’s work and that of many of his successors. Cong & Liu review the relationship between the sentence and the global level in syntactic dependency networks [8] through randomizations of the syntactic structures of the sentences that are used to build a global syntactic dependency network [9]. Cong & Liu argue that the randomization procedure produces non-syntactic global networks, i.e. they take for granted that all what the randomization procedure preserves is purely non-syntactic (notice that the randomization method preserves word frequencies). However, the strong positive correlation between word frequency and word degree, even