<![CDATA[Purpose: This study aimed to construct and evaluate an AI-based Indonesian–Malay bilingual dictionary by incorporating ML and NLP techniques through SLR. This research investigates salient issues in bilingual lexicography at the level of lexis, including limited parallel corpora, contextual semantic variability, and cultural nuances between closely related tongues. Methods: A mixed-methodology, pairing a systematic literature review (SLR) of 18 peer-reviewed studies published from 2020 to 2025 with an experiment based on prototypes. As a result of these findings, we created a mini prototype dictionary via ML-based semantic-mapped supervised and unsupervised models and Transformer representations. We evaluated the prototype using objective measures of translation accuracy and qualitative analysis of errors. Result: NLP and ML techniques are found to be effective in word-level bilingual dictionary building when integrated. Unlike direct equivalence approaches, which neglect contextual understanding in the representation of meaning groups, transformer-based representations capture high degrees of alignment in semantic generality through context. Our prototype produced suitable accuracy for most common lexical entries but struggles with informal expressions due to cultural nuances and underrepresentation of Indonesian–Malay parallel data. Novelty: This study contributes by bridging the existing literature by linking systematic review findings with an empirical implementation in the form of a functional AI-based dictionary prototype, demonstrating that combined machine learning models are effective for Indonesian–Malay lexicography and providing a repeatable framework for AI-driven bilingual dictionaries between other language pairs.]]>
Copyrights © 2026
