<![CDATA[Background: Indonesian–English digital talk increasingly exposes the limits of language technologies that assume stable monolingual input, because switch-points often carry stance, affect, technical meaning, platform practice, and identity across compact social media utterances. Objective: This study examines how Indonesian–English code-switching can be modeled as a computational problem without reducing mixed-language digital utterances to noise, error, or preprocessing residue. Method: Using a corpus-based qualitative-computational design, this study analyzes 15 coded public digital text records through token-level language identification, switch-point mapping, normalization–translation alignment, and pragmatic-computational risk coding. Results: Code-switching appears as structured multilingual practice, with English segments functioning as evaluative phrases, technical labels, platform commands, fandom markers, and discourse-sequence devices within Indonesian frames. Normalization and translation are shown to be interpretive operations, because full correction or translation may flatten sentiment scope, register, and locally meaningful hybrid forms. Monolingual processing risks forced single-language labeling, technical mistranslation, discourse-sequence loss, and representational distortion for multilingual users. Implication: These findings reposition computational adequacy as a question of meaning preservation, not merely model accuracy or benchmark coverage in a low-resource Indonesian digital context for language technologies. Novelty: This study offers an integrated framework that links linguistic switching structure, preprocessing risk, and accountable multilingual NLP representation.]]>
Copyrights © 2026
