Beyond English-only

Towards multilingual instruction with machine translation

Abstract

Australian universities face a critical disconnect between the predominance of monolingual instruction and the multilingual realities of the student population at the tertiary level. With one-fifth of the Australian population speaking a language other than English, existing educational paradigms often fail to foster inclusive learning environments that reflect this linguistic diversity present in the student population. The gap necessitates a fundamental shift in language instruction, which must challenge monolingual conventions and accommodate students from more diverse linguistic backgrounds. While generic translation tools are widely available, they frequently produce fragmented or inaccurate outputs, especially when applied to complex academic content. In the absence of a clear national language policy, this concise paper argues that fine-tuned artificial neural networks used in machine translation, when trained on annotated parallel data, can bridge the equity language instruction gap and facilitate the implementation of multilingual instruction as part of a transformative shift led by educators to ensure no student is left behind.