Spanish–Guaraní Code-Switching in Social Media: A Dependency Parsing Study

Abstract

This study examines how Spanish–Guaraní bilinguals structure code-switching (CS) on Twitter/X and investigates what these patterns reveal about competing models of bilingual grammar. Specifically, it explores language dominance, switch direction, syntactic position, switch span, and register effects in Paraguayan bilingual discourse. The study adopts a corpus-based and dependency-analytic approach using Universal Dependencies (UD) annotation to examine CS patterns in a Spanish–Guaraní social media corpus. The analysis is based on an extended UD-annotated Spanish–Guaraní Twitter/X corpus. Switch points were examined according to direction, dependency relations (DEPREL), part-of-speech categories, span length, and register (formal vs. informal tweets). Quantitative analyses were complemented by qualitative examination of representative examples. Results show that Guaraní, including the variant Jopará, functions as the dominant matrix language overall, especially in formal tweets. Switches into Spanish occur disproportionately at core grammatical positions such as subjects, objects, and obliques, whereas switches into Guaraní frequently occur at clause-edge and discourse-related positions. Switch spans are generally short but vary systematically according to syntactic environment, with clause-structuring elements licensing longer spans than argument-level switches. The study introduces switch span as a measure of structural scope in CS and combines dependency-based syntactic analysis with register-sensitive investigation in a low-resource bilingual setting. The findings support Matrix-Language Frame predictions in formal registers while demonstrating the importance of hierarchical and scope-based accounts of bilingual grammar. More broadly, the study shows how dependency annotation can provide fine-grained insights into CS in Indigenous and mixed-language repertoires.