Simplification in interpreting: the text classification of spoken and interpreted Chinese through ensemble learning techniques

Simplification is a feature of translation universals that has been extensively explored in translation studies. In relation to interpreting, however, recent studies have predominantly focused on a single level and on languages from the Indo-European family, which raises the question of whether the simplification hypothesis holds across different language levels and pairs. In this study, we investigate the simplification hypothesis by using the machine learning technique of ensemble learning to perform a classification experiment comparing interpreted and spoken Chinese. Specifically, we calculate the linguistic complexity borrowed from multiple disciplines, evaluate the characteristics of the two language modes, and apply ensemble learning for classification. We find that interpreted Chinese can be distinguished from spoken Chinese at lexical, syntactic, and combined levels, with classification accuracies of 89.39%, 98.94%, and 99.20%, respectively. However, our further finding that interpreted Chinese exhibits lexical simplification and syntactic complexification suggests a dynamic interplay between syntactic and lexical complexity.