Empirical Translation Studies have recently extended the scope of research to other forms of constrained and mediated communication, including bilingual communication, editing, and intralingual translation. Despite the diversity of factors accounted for so far, this new strand of research is yet to take the leap into intermodal comparisons. In this paper we look at Lexical Diversity (LD), which under different guises, has been studied both within Translation Studies (TS) and Second Language Acquisition (SLA). LD refers to the rate of word repetition, and vocabulary size and depth, and previous research indicates that translated and non-native language tends to be less lexically diverse. There is, however, no study that would investigate both varieties within a unified methodological framework. The study reported here looks at LD in spoken and written modes of constrained and non-constrained language. In a two-step analysis involving Exploratory Factor Analysis and linear mixed-effects regression models we find interpretations to be least lexically diverse and written non-constrained texts to be most diverse. Speeches delivered impromptu are less diverse than those read out loud and the non-constrained texts are more sensitive to such delivery-related differences than the constrained ones.
Binary Classifier Calibration Using an Ensemble of Near Isotonic Regression Models
