A corpus-linguistic account of the history of the genitive alternation in Singapore English

Abstract

In this paper, we are exploring the history of the genitive alternation (of-vs. s-genitive) in Singapore English based on corpus data covering both British English (as the historical input variety) and Singapore English (as the target variety whose diachronic development we are interested in). Specifically, while earlier research has produced partly diachronic accounts of genitive variability, the diachronic development of the genitive has so far not been studied in ESL contexts, a gap which this study attempts to fill. Nearly 6000 instances of of-and s-genitives were annotated for a large number of predictors including phonetic variables (e.g. final sibilancy of possessor), semantic variables (e.g. animacy of possessor/possessum), syntactic variables (e.g. length of possessor/possessum), and pragmatic variables (e.g. discourse accessibility of possessor/possessum). We then applied the method of Multifactorial Prediction and Deviation Analysis with Regressions/Random Forests to the data to explore (i) how genitive choices in Singapore English differ from those in British English and, after a methodological interlude, (ii) how genitive choices changed over time in Singapore English. We conclude with some important recommendations regarding diachronic studies of structural nativization and their theoretical implications in models such as those of Moag (1982) or Schneider (2003, 2007).

Recent Grammatical Change in Postcolonial Englishes: A Real-time Study of Genitive Variation in Caribbean and Indian News Writing

“…In order to look into these similarities and dissimilarities, we opted for Multifactorial Prediction and Deviation Analysis with Regressions (MuPDAR). MuPDAR was developed primarily for learner corpus research (e.g., Gries & Deshors 2014; Wulff & Gries 2015) but has also been applied to the comparison of postcolonial Englishes, particularly to the question of degrees of nativization as well as to the idea of linguistic epicenters (e.g., Gries & Bernaisch 2016; Heller, Bernaisch & Gries 2017; Gries, Bernaisch & Heller 2018). We employed the non-parametric variant of this approach, which is again based not on fixed- or mixed-effects modeling but on random forest analysis, i.e., MuPDAR-F.…”

“…For the reasons just sketched, diachronic investigations have figured prominently in genitive research, but like research on other grammatical changes in English, these investigations have so far focused on the metropolitan varieties and, to a lesser extent, “native” postcolonial Englishes (e.g., Hinrichs & Szmrecsanyi 2007; Hundt & Szmrecsanyi 2012; Szmrecsanyi et al 2017). A few synchronic studies (Szmrecsanyi, Grafmiller, Heller & Röthlisberger 2016; Heller, Bernaisch & Gries 2017; Heller, Szmrecsanyi & Grafmiller 2017; Heller & Szmrecsanyi 2019) have involved other postcolonial Englishes, but to our knowledge there is only a single study that has investigated the development of genitive variation in such a variety based on real-time corpus data: Gries, Bernaisch, and Heller (2018) on Singapore English. In what follows, we present a diachronic analysis of the genitive alternation in Caribbean (here: the Bahamas and Jamaica) and Indian Englishes.…”

On models and modelling

“…Gries et al.’s (2018, p. 249) claim that Mair's (2013) model allows for relatively straight‐forward operationalization and quantification thus underestimates the complexity of the model, especially if hierarchical models of WE try to take the diachronic dimension into account. Statistically testing Mair's (2013) model thus faces similar issues to those encountered in the empirical verification of the epicentre hypothesis (Hundt, 2013).…”

Proceedings of the 2nd International Workshop on Computational Approaches to Historical Language Change 2021

“…In the field of corpus linguistics, large quantities of data have made it possible to precisely model complex multifactorial processes of linguistic change (e.g. Perek and Hilpert, 2017;Gries et al, 2018). Modern methods in natural language processing also increasingly make use of word embeddings, which encode rich information about the use of a word learned from large datasets (Collobert et al, 2011;see Kutuzov et al, 2018 for diachronic word embeddings).…”