The implementation of AI-based image recognition application in slang translation to develop a project-based learning model

Abstract

The advent of AI-powered machine translation technology, particularly text-capture tools such as Google Lens, necessitates that younger generations develop competency in utilizing these applications. This study explores the integration of such technology into translation pedagogy through a digital project-based learning (PjBL) model, designed to enhance translation skills within an interactive educational framework. Translation performance was evaluated using a rubric assessing three key criteria: accuracy, acceptability, and readability. The research objectives were twofold: (1) to evaluate the quality of slang translation produced by students using Google Lens, and (2) to assess students’ understanding and application of this AI tool within a PjBL framework. A mixed-methods approach was employed, following a define-design-develop-disseminate (4D) model. Data were collected through validated tests administered to 75 English education students, organized into five groups. The findings indicate that the overall translation quality of slang terms was poor, with six out of ten target words receiving low scores. A moderate quality level was achieved for two words, and good quality for another two. Conversely, students demonstrated a strong operational understanding of the AI application, reporting familiarity and proficiency in its use for text capture and translation. This discrepancy highlights a gap between technical tool mastery and the critical application required for producing contextually appropriate translations, particularly of informal language like slang.