Determining translation invariant characteristics of James Joyce’s Dubliners

Abstract

We provide a comparative stylometric analysis of the Dubliners stories of James Joyce by using its original and Murat Belge’s Turkish translation. We divide the stories into four categories as suggested by Belge and investigate the success of automatic classification by using discriminant analysis with various style markers. We show that different style markers show different categorization success rates and most of the style markers provide better classification rates in English. We also investigate the sentence, token and type length in both languages. We show that sentence lengths are linearly mapped from English to Turkish, type and token length distribution follow the Poisson distribution in both languages, and the related relative frequency curves provide us with an invariant between the original text and the translation