Patterns in syntactic dependency networks from authored and randomised texts

The syntactic relationships between words allow a communicator to express a virtually endless array of thoughts by a finite set of elements. The co-occurrence of words in a sentence reflects the syntactic dependency between words, and can be represented as a directed graph. In this account we compiled the grammar dependency networks of 86 texts from 11 well known English authors. In an analysis of the common and specific features of these networks we try to attribute network properties to individual authors. A pointwise defined measure shows no significant groups which could be identified with authors. Further, a comparison to randomized versions of the same texts shows a systematic, but very small difference between networks constructed for the originals and the randomisations, respectively. This suggests, that the scale-free and small world-like nature of these networks can be explained by an underlying regularity in the word frequency distribution, known as Zipf’s law. A stochastic model, which allows the construction of networks for arbitrary word frequency distributions, illustrates this idea.