Lexical diversity as a lens into the classification of Slavic languages: A quantitative typology perspective

This study proposes a linguistic classification method based on quantitative typology, which leverages a large-scale multilingual parallel corpus to obtain valid language classification result by excluding the influence of covariates such as text genre and semantic content in cross-language comparison. To achieve this, we model the type–token relationships of each Slavic parallel text and calculate the lexical diversity to approximate the morphological complexity of the language. We perform automatic clustering of languages based on these lexical diversity metrics. Our findings show that (1) the lexical diversity metrics can well reflect that the language is located somewhere on the continuum of ‘analytism-synthetism’; (2) the automatic clustering based on these metrics effectively reflects the genealogical classification of Slavic languages; and (3) the geographical distribution of lexical diversity in the region where Slavic languages are spoken shows a monotonic increasing trend from southwest to northeast, which is consistent with the pattern found by previous authors on a global scale. The methodological approach taken in this study is data-driven, with the benefit of being independent of theoretical assumptions and easy for computer processing. This approach can offer a better insight into corpus-based typology and may shed light on the understanding of language as a human-driven complex adaptive system.