Mapping Technology-Enhanced Language Assessment in TVET and Occupational Education: Bibliographic Structure and Emerging Directions
Abstract
Technology-enhanced language assessment increasingly combines online testing, automated scoring, artificial-intelligence-supported feedback and multimodal performance capture. In vocational and occupational education, however, this literature is dispersed across language testing, English for specific purposes, workplace communication and professional education. This study maps the bibliographic structure of that intersection and examines what the underlying studies show about occupational authenticity, human-AI judgement, validity evidence and assessment governance. A Scopus search conducted on 5 August 2026 retrieved 253 records. After document-type, language, publication-year and title-abstract screening, 31 journal articles meeting all three substantive eligibility dimensions formed the final bibliometric corpus. Performance analysis, a reproducible Python/NetworkX document-level bibliographic-coupling analysis, author-keyword analysis and structured full-text synthesis were combined. Twenty-two articles (71.0%) were published from 2021 to 2026. At a threshold of at least one shared cited work, the coupling network contained 44 unique document-pair links with a summed edge weight of 65; the largest component included 19 documents and 12 documents were isolated. The same largest-component size and isolate count were retained across exact, prefix and fuzzy canonicalisation variants, although link counts and summed edge weights varied. Author-keyword recurrence was sparse, with only six terms occurring in at least two documents. Full texts were available for 26 articles and supported four cross-cutting findings: human-AI assessment is more defensible as complementarity than substitution; occupational authenticity depends on task-construct alignment; validity evidence remains uneven across assessment uses; and governance and score-use evidence are less developed than delivery, scoring and feedback. The field is therefore recent but structurally fragmented. Future research should prioritise occupationally authentic validation, transparent human-AI decision rules, fairness, consequential score use and traceable assessment governance.