Speaker
Description
Lexical blending is a highly productive source of new English vocabulary, and EFL learners now encounter items such as staycation, infodemic, and covidiot largely through AI-mediated platforms that supply on-demand explanations of their formation. However, large language models generalise morphological patterns unreliably and can produce fluent but inaccurate word-formation accounts (Ismayilzada et al., 2024), while leading AI-literacy frameworks (Long & Magerko, 2020; Ng et al., 2021) remain cross-domain and do not operationalise the evaluative capacities vocabulary learners require. This conceptual paper aims to develop and theorise NACPE (Notice–Analyse–Categorise–Produce–Evaluate), a corpus- and AI-integrated framework for building vocabulary-specific AI literacy through morphological awareness. The framework was constructed through a structured synthesis of three literatures :corpus-linguistic evidence that blending is rule-governed (Gries, 2004; Beliaeva, 2019; Mattiello, 2019), L2 research linking morphological awareness to lexical inferencing (Zhang & Koda, 2012), and AI-literacy scholarship and it mapped onto a five-stage pedagogical sequence and illustrated through worked classroom tasks drawing on COCA and the NOW corpus. The synthesis yields three outcomes: a set of domain-specific criteria for evaluating AI-generated word-formation explanations; prosodic and structural well-formedness tests learners can apply independently; and a transferable task template for corpus-based verification. The paper contributes a replicable model for integrating data-driven learning with critical AI engagement, advancing learner agency and assessment literacy in technology-mediated vocabulary instruction.
Keywords: AI literacy, morphological awareness, lexical blending, data-driven learning, EFL vocabulary
