This study focuses on addressing the challenges associated with applying Large Language Models (LLMs) and the Retrieval Augmented Generation (RAG) method, while also incorporating voice processing to resolve students' inquiries about information for the school in areas such as admissions, academic advising, scientific research, and updates on new information and announcements from the International School – Vietnam National University, Hanoi. Although LLMs have made significant advancements in natural language processing, they still encounter difficulties when handling specific applications. The objective of this research is to enhance the knowledge and reasoning capabilities of LLMs to address issues that require student consultation. The proposed approach includes fine-tuning the Large Language Model, utilizing Retrieval Augmented Generation combined with training high-performance retrieval models such as Embedding models, Reranking models, and BM25. Additionally, I will fine-tune Text-to-Speech and Speech-to-Text models to handle audio issues using a proprietary dataset. This research aims to improve the accuracy and efficiency of RAG-based approaches in supporting the advisory team of the International School. These findings have the potential to contribute to the advancement of educational technology, enabling students, teachers, and users to access information about the International School easily and accurately.
Readership Map
Content Distribution
This study focuses on addressing the challenges associated with applying Large Language Models (LLMs) and the Retrieval Augmented Generation (RAG) method, while also incorporating voice processing to resolve students' inquiries about information for the school in areas such as admissions, academic advising, scientific research, and updates on new information and announcements from the International School – Vietnam National University, Hanoi. Although LLMs have made significant advancements in natural language processing, they still encounter difficulties when handling specific applications. The objective of this research is to enhance the knowledge and reasoning capabilities of LLMs to address issues that require student consultation. The proposed approach includes fine-tuning the Large Language Model, utilizing Retrieval Augmented Generation combined with training high-performance retrieval models such as Embedding models, Reranking models, and BM25. Additionally, I will fine-tune Text-to-Speech and Speech-to-Text models to handle audio issues using a proprietary dataset. This research aims to improve the accuracy and efficiency of RAG-based approaches in supporting the advisory team of the International School. These findings have the potential to contribute to the advancement of educational technology, enabling students, teachers, and users to access information about the International School easily and accurately.
