Cross-Lingual Transfer Entropy Mapping in Pragmatic Inference: A Bayesian Discourse-Analytic Framework for Second Language Acquisition Diagnostics

Abstract

The cross-lingual transfer of pragmatic inference mechanisms remains a structurally underspecified domain within second language acquisition (SLA) research, particularly regarding the computational modeling of discourse-level entailment patterns. This study introduces a Bayesian discourse-analytic framework that integrates transfer entropy mapping with corpus-driven pragmatic annotation schemas to diagnose L1-to-L2 inferential drift in advanced multilingual learners. Drawing on a longitudinal corpus of 1,240 annotated spoken and written discourse samples across four typologically divergent language pairs, the proposed framework operationalizes illocutionary force miscalibration as a measurable diagnostic variable. Results demonstrate that entropy-weighted pragmatic transfer indices significantly outperform conventional interlanguage error taxonomies in predictive accuracy. The framework offers scalable diagnostic protocols applicable to both instructed SLA environments and computational language assessment pipelines, advancing the theoretical interface between corpus linguistics, pragmatics, and probabilistic modeling.