Possible Chaotic Structures in the Turkish Language with Time Series Analysis Avadis Hacınlıyan 1 , Murat Erentürk 1 , and Gkhan Şahin 1 1 Department of Physics and Department of Information Technologies, Yeditepe University, İstanbul, Turkey {ahacinliyan, merenturk, sahin}@yeditepe.edu.tr Abstract The possibility of chaotic structures in Turkish and English texts, as well as the possibility of using the pseudo-invariants in a reconstructed phase space as identifying characteristics for languages is investigated. Texts of length up to 83000 in both languages have been analysed. Two alternatives for the dependent variable in a time series analysis have been used. Word frequencies based on a corpus have been one alternative inspired by Zipfs law. The other alternative is based on assigning values to the letters in a word as inspired by a random walk. A positive maximal lyapunov exponent has been observed. Values of this exponent are different for the two languages. This and differing detrended fluctuation analysis results for the two languages for either parametrization imply that our analysis methods can point to differences in languages. 1. INTRODUCTION The structure of natural languages has recently become an important field of research following the observation that texts written in natural languages obey laws that approximately obey rules of fractal geometries [1,2] . This behavior is characteristic of many other systems in nature including many forms of music. On the other hand, time series analysis methods [3,4] have become an import tool in analysing fractal structures using a one dimensional signal in time. Unfortunately, there are several ways in which one can generate a one dimensional time signal based on a literary text. A natural language is a hierarchy of structures that involve both a sound and a meaning. The simplest structures consist of the letters of the alphabet and the syllables; they will have a contribution to the sound but not to the meaning. Structures higher in the hierarchy such as words, sentences and paragraphs will contribute to the meaning. In order to be an effective communication medium, a language must be patterned. It is expected that the patterns will involve self similarity [2,5] and there should be a certain characteristic distance or distances between words that significantly contribute to the meaning ,[6] . It would be tempting to see time series analysis can give us an estimate on both the possibility of self similarity and the existence of such a window, since chaotic behaviour is related to the long time predictability of a system .