Global Syntactic Variation in Seven Languages: Toward a Computational Dialectology

Abstract

The goal of this paper is to provide a complete representation of regional linguistic variation on a global scale. To this end, the paper focuses on removing three constraints that have previously limited work within dialectology/dialectometry. First, rather than assuming a fixed and incomplete set of variants, we use Computational Construction Grammar to provide a replicable and falsifiable set of syntactic features. Second, rather than assuming a specific area of interest, we use global language mapping based on web-crawled and social media datasets to determine the selection of national varieties. Third, rather than looking at a single language in isolation, we model seven major languages together using the same methods: Arabic, English, French, German, Portuguese, Russian, and Spanish. Results show that models for each language are able to robustly predict the region-of-origin of held-out samples better using Construction Grammars than using simpler syntactic features. These global-scale experiments are used to argue that new methods in computational sociolinguistics are able to provide more generalized models of regional variation that are essential for understanding language variation and change at scale.

Proceedings of the Workshop on Cognitive Modeling and Computational Linguistics

“…The inverse of this generalization is that individuals have unique or idiosyncratic constructions which are only revealed when the training corpus is centered around that individual. This finding fits well with studies in variation (Dunn, 2019b), Dunn2019a which reveal the high degree of syntactic differences across speech com- munities.…”

Mapping languages: the Corpus of Global Language Use

“…Variation within and between both datasets is structured more around individual languages and is less predictable given country-specific population and corpus size information. Work based on previous versions of the corpus (Dunn, 2019a(Dunn, , 2019b have shown that meaningful dialectal variation can be modeled using this source of data. The internal (corpus similarity) and external (demographic) evaluations in this paper strongly suggest that future work based on these expanded country-language sub-corpora will support further advances in corpus-based dialectology.…”

Production vs Perception: The Role of Individuality in Usage-Based Grammar Induction

“…The grammar induction algorithm used here employs an association-based beam search to identify the best sequences of slot-constraints (Dunn, 2019a). While a grammar formalism like dependency grammar (Nivre and McDonald, 2008;Zhang and Nivre, 2012) must identify the head and attachment type for each word, a construction grammar must identify the representation type for each slot-constraint.…”

“…However, because these two types of representations operate at different levels of complexity, it is possible that they grow at different rates. We thus experiment with the growth of a computational construction grammar (Dunn, 2018b(Dunn, , 2019a across data drawn from six different registers: news articles, Wikipedia articles, web pages, tweets, academic papers, and published books. These experiments are needed to establish a baseline relationship between the grammar and the lexicon for the experiments to follow.…”