Change of Word Characteristics in 20th-Century Turkish Literature: A Statistical Analysis

This paper provides a century-wide quantitative analysis of the Turkish literature using forty novels of forty authors. We divide the century into four eras or quarter centuries; allocate ten novels to each era, and partition each novel into equal sized blocks. Using cross validation based discriminant analysis, with the most frequent words as discriminators; we achieve a classification rate with a relatively high accuracy when the novel blocks are classified according to their eras. We show that, by using statistical stylistic methods, the author gender of Turkish text can be accurately identified. We also study the gender differences regarding the use of most frequent words. Using weighted least squares regression and a sliding window approach we show that as time passes, words, both in terms of tokens (in text) and types (in vocabulary), have become longer. The findings of this work have implications for the historical linguistic analysis of the Turkish language. Abbreviated title: Change of Word Characteristics in Turkish Corresponding author, voice: +90 (312) 290-2613, fax: +90 (312) 266-4047. CHANGE OF WORD CHARACTERISTICS IN TURKISH P. 2