Design and Development of Meta-Heuristic Optimized Generative Models for Enriching English Machine Translation

With the advent of Generative Artificial Intelligence (GAI) and Large Language Models (LLM), new flavour of teaching by replacing the humans has emerged in the recent educational scenarios. However, low-resource contents in cross- languages and high complex training are considered to be major bottleneck for an effective implementation of Artificial Intelligence (AI) based teaching agents for fulfilling the language barriers. In this context, this research article aims to develop the hybrid generative AI framework that utilizes finetuned language model by the meta-heuristic technique to achieve the high translation accuracy. The suggested hybrid framework incorporates the fine-tuned Mistral-7B model integrated with the meta-heuristic algorithm for a achieving the high rate of translated contents in accordance to the student's known language. The comprehensive experimentation is carried out using the different multi-lingual datasets and various metrics such as BLEU, ChF and WER are analysed and validated with the other language models such as LLAMA, GEMMA, GPT.3.0 and BERT. The complete data -driven outcomes demonstrated that the proposed meta-heuristic model has achieved BLEU score of 0.9, chF score of 0.99 and WER with 0.01, which outshines the other models in producing the enhanced reliability of produced contents. These findings highlights the potential of hybrid large language models for generating the more reliable contents, offering the more scalable system for breaking the language barrier