Statistical and Deep Learning Methods for a Linguistic and Literary Analysis

This chapter presents a study of characteristic morphosyntactic elements of Milan Kundera’s production, which are detected by linguistic, statistical and machine learning approaches. The specificity of this contribution is to propose, in addition to the traditional statistical methods, a deep learning training on a database comprising Kundera’s essays and novels and a representative sample of French contemporary authors, using grammatical categories as sole representation of the texts. This study led to the identification of a morphosyntactic pattern, which was then examined to reveal the aesthetic intention behind its use.