The AMTA Framework: An AI-Augmented Triangulation Model for Empirical Bilingual Lexicography

This study introduces and empirically validates the AI-augmented Method of Triangulation Approach (AMTA), a novel framework designed to advance specialized bilingual lexicography by systematically integrating corpus linguistics, artificial intelligence (AI), and expert validation. The methodology operationalizes a tripartite process comprising: (1) a hybrid term extraction protocol coupling nested-term adjustment (C-Value logic) with log-likelihood (G 2 ) keyness calculations against the British Academic Written English (BAWE) corpus to isolate valid multiword units (MWUs); (2) AI-augmented content generation via a Retrieval-Augmented Cascade Architecture (RACA) that enforces a fallback protocol privileging academic contexts to mitigate hallucinatory risks, enhanced by retrieval-augmented generation (RAG); and (3) a multidimensional validation layer incorporating quantitative metrics and expert review. Implemented on a scalable Google Apps Script platform that leverages FREE-tier Application Programming Interfaces (APIs), the framework compiled an English-Arabic dictionary of AI terminology in education. The initial results demonstrate that the AMTA model effectively bridges computational efficiency with lexicographic rigor, successfully processing 517 domain-salient terms and yielding 234 expert-validated entries with high ratings for definitional accuracy and pedagogical utility. This study establishes a reproducible, empirically grounded model that offers a viable and scalable pathway for institutions with limited resources, contributing a significant methodological advance in the development of dynamic, data-driven, and reliable bilingual dictionaries for rapidly evolving specialized domains.