Optimization of data analysis models for low-resource Eurasian languages using machine translation

Abstract

This study explores low-resource language data translation models in the realms of multimedia teaching and cyber security. A rapid learning-based neural machine translation (NMT) method is developed based on meta-learning theory. Subsequently, the back translation method is employed to further improve the NMT model for low-resource language data. Results indicate that the proposed low-resource language NMT method based on meta-learning achieves increased Bilingual Evaluation Understudy (BLEU) scores for three target tasks in a supervised environment. This study emphasizes the auxiliary role of meta-learning theory in low-resource language data translation, aiming to enhance the efficiency of translation models in utilizing information from low-resource languages.