Conceptual Network Analysis of Artificial Intelligence in Language Education

This study maps the conceptual landscape of artificial intelligence (AI) research in language education through a systematic conceptual network analysis of 98 peer-reviewed journal articles retrieved from Web of Science (2020-2026). Drawing on Positive Pointwise Mutual Information (PPMI) weighting, Louvain community detection, and multi-index centrality computation, the analysis identifies 17 validated concepts, 58 significant co-occurrence edges, and three thematic communities. Learner Motivation emerges as the most structurally central concept despite belonging to the underrepresented Educational Outcomes dimension, while Natural Language Processing and Machine Learning function as cross-community bridges. Large Language Models, though the most frequently cited concept, display comparatively low network centrality, suggesting that this rapidly growing research strand remains insufficiently integrated with established pedagogical and linguistic frameworks. The findings reveal a pronounced technology-to-pedagogy imbalance and point to concrete directions for future research that foreground learning effectiveness alongside technological capability.