Modeling language evolution and feature dynamics in a realistic geographic environment

Proceedings of the 28th International Conference on Computational Linguistics, pages 788–798 Barcelona, Spain (Online), December 8-13, 2020 788 Modeling language evolution and feature dynamics in a realistic geographic environment Rhea Kapur The Pingry School rhea.kapur1@gmail.com Phillip Rogers Department of Linguistics The University of California, Santa Barbara phillip@ucsb.edu Abstract Recent, innovative efforts to understand the uneven distribution of languages and linguistic fea- ture values in time and space attest to both the challenge these issues pose and the value in solving them. In this paper, we introduce a model for simulating languages and their features over time in a realistic geographic environment. At its core is a model of language phylogeny and migration whose parameters are chosen to reproduce known language family sizes and geographic disper- sions. This foundation in turn is used to explore the dynamics of linguistic features. Languages are assigned feature values that can change randomly or under the influence of nearby languages according to predetermined probabilities. We assess the effects of these settings on resulting geographic and genealogical patterns using homogeneity measures defined in the literature. The resulting model is both flexible and realistic, and it can be employed to answer a wide range of related questions.1 1 Introduction The distribution of linguistic feature values across time and space is uneven. The overall frequencies of particular feature values can be skewed with some highly frequent and others vanishingly rare, and many feature values have been observed to cluster geographically or within particular language families. Such patterns are most commonly explained in terms of the competing influences of inheritance, diffusion, universal tendencies of language drift, and chance. Understanding these dynamics has long been a goal of linguists, and it has important implications for historical linguistics, areal linguistics, and linguistic ty- pology. Despite growing attention to these issues and advances in comparative and quantitative methods, many questions remain unanswered. The well-known problem of spatial autocorrelation limits the usefulness of a direct statistical ap- proach to disentangling geography from genealogy, and as a result various other approaches have been innovated. There are a number of studies exploring the stability or biases of typological features within language families (Dediu and Cysouw, 2013; Dediu and Levinson, 2012; Maslova, 2000; Wichmann and Holman, 2009). Of these approaches, those that account for geography are limited to comparisons within families or between families. Other successful approaches which account for both genealogy and geography are limited to a particular region (Cysouw, 2013), use typological distance measures that do not allow for the comparison of features separately (e.g. Comrie and Cysouw, 2012), or apply a ratio of genealogical and geographic measures that reduces independently varying factors to their quotient (Parkvall, 2008). Computer simulations have also emerged as a promising method by which to explore these and related issues (e.g., Holman et al., 2007; Oliveira et al., 2008). However, nearly all of these models employ abstract representations of space, such as a lattice. A recent exception is Wichmann (2017), which uses ‘populated places’ from the GeoNames database to …

Content truncated to 500 words