Does network complexity help organize Babel’s library?

Abstract

Quantitative linguistics has contributed significantly to the study of the origins and properties of natural languages. Among many statistical properties described, one of the most basic but no less fundamental, is the well-known Zipf's law. Its ubiquity would reveal underlying principles of natural languages functioning. Although the presence of this law reduces drastically the probability that a text can be the result of a random process, it is insufficient information to determine if the text is the result of a ciphering process that shifts the position of words of a "readable" text, or if the text corresponds to a non-sense one constructed with a "bag of words" with a Zipfian distribution. In this work we show that simple global topological properties of co-ocurrent word networks constructed from texts, seem to be the fingerprint of the sense texts. We observe that many statistical properties of these networks depend on the frequency of words in the text, however, others seem to be strictly determined by the grammar. Our results suggest that seems to be a lower bound of sense that depends on the correlation between mean word connectivity and word connectivity correlation. This property, in addition to being only present in sense text, and absent in, until now, not decoded texts such as Voynich Manuscript, would also be exclusive for natural languages, allowing us to discriminate between these and formal texts.keywords: complex networks, word networks,Voynich manuscript, quantitative linguistics

Consilience of Reductionism and Complexity Theory in Language Research: Adaptive Weight Model

“…Recently, the machine learning model has demonstrated the potential to solve this problem. Machine learning has been proved to be able to simulate the language processing process and solve issues of language research [4]. Machine learning algorithms embrace interconnectedness and dynamism [5], which embodies the philosophy of complexity theory.…”