Modeling the Effectiveness of an AR-Based Context-Aware Platform for English Translation Learning

Abstract

Traditional English translation instruction has long faced persistent challenges, including the absence of authentic contexts, delayed feedback, and insufficient personalization. With the advancement of educational technologies, mobile learning platforms that integrate augmented reality (AR) and context-aware mechanisms have opened new pathways for creating immersive and situational translation learning environments. However, existing studies have primarily concentrated on technological implementation and user satisfaction, while providing limited quantitative evidence regarding the key factors that determine instructional effectiveness. In response to this gap, a modeling-based investigation was conducted to examine the mechanisms through which an integrated AR and context-aware mobile interactive platform influences translation learning outcomes. A prototype platform was designed and developed to deliver visual contextual support through AR and to dynamically adapt learning content via context-aware sensing. Through instructional experiments, pre- and post-test translation performance data, platform interaction logs, and contextual information were systematically collected. A random forest regression model was employed, with translation competence improvement as the predictive target, to conduct multidimensional feature modeling and analysis. The results demonstrated that the experimental group achieved significantly greater improvements in overall translation competence, as well as in linguistic accuracy, pragmatic appropriateness, and contextual adaptability, compared to the control group, confirming the instructional effectiveness of the AR-based context-aware platform. Moreover, the random forest model exhibited superior predictive accuracy compared with traditional models. AR scene interaction frequency and contextual complexity were found to be the primary driving features, showing a positive interaction effect—enhanced benefits were observed under conditions of high contextual complexity combined with deep AR interactions.